{"page-breadcrumb.c766e66518":"Breadcrumb","trust-centre.breadcrumb.home":"Home","trust-centre.breadcrumb.trust":"Trust Centre","trust-centre.shared.publicRecord":"Public record","trust-centre.shared.updated":"Updated 1 August 2026","trust-centre.nav.label":"Trust Centre records","trust-centre.nav.hub":"Overview","trust-centre.nav.transparency":"AI transparency","trust-centre.nav.models":"Models","trust-centre.nav.loom":"Loom record","trust-centre.nav.training":"Training data","trust-centre.nav.generated":"Output marking","trust-centre.nav.use":"Acceptable use","trust-centre.nav.terms":"Commercial terms","trust-centre.nav.safety":"Safety and security","trust-centre.nav.evaluations":"Evaluations","trust-centre.nav.risk":"Risk","trust-centre.nav.monitoring":"Monitoring","trust-centre.nav.incidents":"Incidents","trust-centre.nav.data":"Data governance","trust-centre.nav.documentation":"Technical records","trust-centre.nav.copyright":"Copyright","trust-centre.nav.downstream":"Downstream","trust-centre.nav.report":"Report","trust-centre.nav.changes":"Changes","trust-centre.status.verified":"Documented|||Internally confirmed against a current Dweve record, implementation or owner-provided fact as of the displayed date; not a claim of independent audit or certification","trust-centre.status.prepared":"Pre-release|||Built or prepared for the planned external release","trust-centre.status.pending":"Future event|||The method and evidence contract are current; the named real-world event has not yet occurred","trust-centre.status.verified.short":"Documented","trust-centre.status.prepared.short":"Pre-release","trust-centre.status.pending.short":"Future event","trust-centre.status.boundary.short":"Boundary","trust-centre.next.eyebrow":"Public channel","trust-centre.next.title":"A record should be challengeable","trust-centre.next.body":"Ask for missing information, report a concern, challenge a copyright decision, or tell us where a public record is unclear. The message enters the same recorded contact route.","trust-centre.next.action":"Open the reporting desk","trust-centre.page.hub.crumb":"Overview","trust-centre.page.hub.hero.eyebrow":"Dweve Trust Centre","trust-centre.page.hub.hero.title":"Trust is not a promise|||it is built into the system","trust-centre.page.hub.hero.lede":"The public evidence behind Dweve Loom, its data route, operating boundaries and release controls. Documented means internally confirmed as of the displayed date; independent assurance is named separately when it exists.","trust-centre.page.hub.hero.actions":"Inspect the public record|||Open the Loom record","trust-centre.page.hub.glyph":"Public evidence register|||Pre-release","trust-centre.page.transparency.crumb":"AI transparency","trust-centre.page.transparency.hero.eyebrow":"AI governance","trust-centre.page.transparency.hero.title":"The commitment is public|||the evidence stays up to date","trust-centre.page.transparency.hero.lede":"Dweve is a voluntary signatory to the EU General-Purpose AI Code of Practice and participates in the Signatory Taskforce. This dossier explains what that commitment means in Dweve's architecture and where the evidence is kept.","trust-centre.page.transparency.hero.actions":"Read our position|||Inspect the change record","trust-centre.page.transparency.glyph":"Governance position|||Active","trust-centre.page.models.crumb":"Model registry","trust-centre.page.models.hero.eyebrow":"Public model registry","trust-centre.page.models.hero.title":"One model family|||one accountable record","trust-centre.page.models.hero.lede":"Dweve Loom 1.0 is the only Dweve model. The registry separates its adaptive model identity, captured states, access routes and factual release state without treating Loom as one static hash.","trust-centre.page.models.hero.actions":"Open the registry|||Read the Loom record","trust-centre.page.models.glyph":"Model registry|||One entry","trust-centre.page.loom.crumb":"Dweve Loom 1.0","trust-centre.page.loom.hero.eyebrow":"Model record / Dweve Loom 1.0","trust-centre.page.loom.hero.title":"A model is one thread|||Loom is the weave","trust-centre.page.loom.hero.lede":"Loom is built around constraint learning and a typed execution graph. It combines perception, memory, retrieval, domain specialists, solvers, and verification. This record states what changes, what is sealed, and what can be replayed.","trust-centre.page.loom.hero.actions":"Read the architecture|||Inspect training content","trust-centre.page.loom.glyph":"Loom model record|||Internal testing","trust-centre.page.training.crumb":"Training content","trust-centre.page.training.hero.eyebrow":"Loom training-content summary","trust-centre.page.training.hero.title":"Every source family is named|||every admission still has to prove itself","trust-centre.page.training.hero.lede":"The register publishes 772 supplied training and knowledge-source entries across 19 groups. Each entry states its licence gate, usable slice, and purpose. Exact releases enter Loom 1.0 only after their rights, provenance, privacy, and evidence checks pass.","trust-centre.page.training.hero.actions":"Inspect the source register|||Read the copyright policy","trust-centre.page.training.glyph":"Training register|||772 sources · current","trust-centre.page.copyright.crumb":"Copyright and crawling","trust-centre.page.copyright.hero.eyebrow":"Copyright and source access","trust-centre.page.copyright.hero.title":"Collection starts with rules|||not with a fetch","trust-centre.page.copyright.hero.lede":"Winnow checks source policy, robots.txt, rates, targets, and request records. Spindle governs provenance and promotion. Loom consumes governed material and can use Winnow as a tool.","trust-centre.page.copyright.hero.actions":"Inspect the controls|||Contact the rights desk","trust-centre.page.copyright.glyph":"Copyright control record|||Public channel open","trust-centre.page.downstream.crumb":"Downstream documentation","trust-centre.page.downstream.hero.eyebrow":"For integrators and deployers","trust-centre.page.downstream.hero.title":"The model record travels|||with the integration","trust-centre.page.downstream.hero.lede":"Downstream teams need current capabilities, limits, interfaces, evaluation context, and changes. Dweve publishes the common record first and scopes private details only where necessary.","trust-centre.page.downstream.hero.actions":"Inspect the package|||Request more information","trust-centre.page.downstream.glyph":"Downstream information|||Prepared","trust-centre.page.generated.crumb":"Output transparency (Article 50)","trust-centre.page.generated.hero.eyebrow":"Output transparency · Article 50","trust-centre.page.generated.hero.title":"Every release carries|||a verifiable origin","trust-centre.page.generated.hero.lede":"Dweve's release-bound policy covers Loom text and code output plus the image output Dweve ships. Loom does not generate video. C2PA Content Credentials carry the machine-readable provenance record, and route-specific validation runs before an external route opens.","trust-centre.page.generated.hero.actions":"Inspect the marking policy|||Report a model issue","trust-centre.page.generated.glyph":"Output transparency|||Release-bound controls","trust-centre.page.use.crumb":"Acceptable use","trust-centre.page.use.hero.eyebrow":"Acceptable-use boundary","trust-centre.page.use.hero.title":"Capability needs authority|||before action","trust-centre.page.use.hero.lede":"Dweve tools are built for legitimate, governed work. A technically possible action is not automatically permitted. Purpose, authority, evidence, human responsibility, and deployment policy remain part of the execution boundary.","trust-centre.page.use.hero.actions":"Read the boundaries|||Report unsafe use","trust-centre.page.use.glyph":"Use boundary|||Policy prepared","trust-centre.page.report.crumb":"Reporting desk","trust-centre.page.report.hero.eyebrow":"One public reporting desk","trust-centre.page.report.hero.title":"Send the concern once|||we route the record","trust-centre.page.report.hero.lede":"Copyright, downstream information, model safety, security, privacy, and model-quality concerns enter one recorded route. Choose the topic, preserve useful evidence, and receive a submission reference.","trust-centre.page.report.hero.actions":"Choose a reporting route|||Return to the Trust Centre","trust-centre.page.report.glyph":"Reporting desk|||Receiving","trust-centre.page.changes.crumb":"Change record","trust-centre.page.changes.hero.eyebrow":"Trust record history","trust-centre.page.changes.hero.title":"No silent revisions|||no missing history","trust-centre.page.changes.hero.lede":"Material changes to the model record, source policy, public controls, and machine-readable files are listed here with their effect. Human and machine feeds carry the same version.","trust-centre.page.changes.hero.actions":"Read the current release|||Open the JSON record","trust-centre.page.changes.glyph":"Trust change ledger|||Version 1.0","trust-centre.fact.signatory":"GPAI Code signatory|||Dweve signed voluntarily and publishes the implementation path","trust-centre.fact.taskforce":"Signatory Taskforce|||Dweve participates in the working forum for signatories","trust-centre.fact.release":"External access|||Invite-only external access is planned for 1 September 2026","trust-centre.fact.training":"Training source register|||772 supplied source entries are public; exact release admissions remain evidence-bound","trust-centre.fact.voluntary":"Full voluntary scope|||Transparency, copyright, safety, and security measures are included","trust-centre.hub.records.eyebrow":"Public record map","trust-centre.hub.records.heading":"Every claim has|||a place to inspect it","trust-centre.hub.records.body":"The hub is organised by the question a reader needs answered. Model identity is separate from training content. Copyright controls are separate from acceptable use. Each record exposes its own status.","trust-centre.hub.records.pills":"Model|||Data|||Operation","trust-centre.hub.records.chrome":"How to read the status|||Public evidence map|||Ten records|||Missing fields stay visible","trust-centre.record.transparency":"AI transparency|||Signatory position, voluntary scope, governance, and claim boundaries","trust-centre.record.loom":"Loom model record|||Architecture, modalities, state, replay, and operating data policy","trust-centre.record.training":"Training content|||772 supplied source entries, deterministic admission, and release-specific evidence status","trust-centre.record.copyright":"Copyright and crawling|||Winnow controls, rights route, and machine-readable policy","trust-centre.record.downstream":"Downstream documentation|||The common integration and evaluation information package","trust-centre.record.generated":"Generated content|||Output scope, trace attachment, and marking controls","trust-centre.record.use":"Acceptable use|||Purpose, authority, refusal, review, and appeal boundaries","trust-centre.record.report":"Reporting desk|||One route for copyright, safety, security, privacy, and model concerns","trust-centre.record.changes":"Change record|||Version history in HTML, JSON, and Atom","trust-centre.record.complianceCost":"Compliance cost|||Scope, period, inclusions, and exclusions for the compliance programme","trust-centre.hub.chain.eyebrow":"Evidence chain","trust-centre.hub.chain.heading":"From an outside source|||to an accountable output","trust-centre.hub.chain.body":"Winnow acquires under a declared source policy. Spindle preserves provenance and decides what becomes governed knowledge. Loom uses that state and can call Winnow for current information.","trust-centre.hub.chain.pills":"Acquire|||Govern|||Use","trust-centre.hub.chain.chrome":"Three control layers|||Source-to-output route|||Typed hand-offs|||No hidden scrape path","trust-centre.chain.policy":"Access policy|||Robots, rates, target safety, and operator rules apply before collection","trust-centre.chain.evidence":"Provenance|||Source, address, time, status, and content identity stay attached","trust-centre.chain.runtime":"Runtime boundary|||Loom receives governed material or invokes Winnow through a tool boundary","trust-centre.chain.winnow":"Winnow|||Discovers, fetches, parses, and records according to source rules","trust-centre.chain.spindle":"Spindle|||Assesses provenance, evidence, quality, disagreement, and promotion","trust-centre.chain.loom":"Loom|||Selects the relevant cognitive weave and records what it used","trust-centre.chain.output":"Output record|||The answer carries its trace, sources, state, and execution context","trust-centre.hub.release.eyebrow":"Release state","trust-centre.hub.release.heading":"Pre-release does not mean|||unrecorded","trust-centre.hub.release.body":"Loom is currently in internal-only testing. External invite-only access is planned for 1 September 2026 across individual and combined product routes. The public record makes the remaining release-readiness steps explicit.","trust-centre.hub.release.pills":"Internal now|||Invite-only next|||Versioned after","trust-centre.hub.release.chrome":"Current posture|||Release status line|||Three states|||Dates are not hidden","trust-centre.release.internal":"Internal testing|||Dweve teams operate the current pre-release system internally","trust-centre.release.external":"External invite-only|||Testers, partners, investors, and selected participants enter by invitation","trust-centre.release.continuous":"Post-market record|||Usage, incidents, evaluations, and changes continue after access opens","trust-centre.release.now":"Now|||Internal-only pre-release testing with public records being prepared","trust-centre.release.gate":"Before 1 September|||Complete the external-access controls and keep each admitted item bound to its controlled Spindle record","trust-centre.release.after":"From 1 September|||Invite-only external operation with versioned monitoring and change publication","trust-centre.hub.contact.eyebrow":"Challenge route","trust-centre.hub.contact.heading":"A public record needs|||a public way to challenge it","trust-centre.hub.contact.body":"The reporting desk separates the subject while keeping one submission route. Sensitive material should be described first, not pasted into a public or general-purpose field.","trust-centre.hub.contact.pills":"Rights|||Safety|||Information","trust-centre.hub.contact.chrome":"Reporting posture|||Choose the right desk|||Recorded route|||One submission, clear owner","trust-centre.contact.public":"Public entry point|||Every route starts from the same accessible form","trust-centre.contact.recorded":"Submission record|||Successful delivery returns a reference from the contact system","trust-centre.contact.routed":"Subject ownership|||The selected topic determines the internal owner and evidence request","trust-centre.contact.rightsholder":"Rightsholder request|||Raise a source, reservation, removal, or copyright concern","trust-centre.contact.downstream":"Downstream request|||Ask for integration, evaluation, capability, or limitation information","trust-centre.contact.safety":"Model safety|||Report harmful behavior, unexpected capability, or a failed control","trust-centre.contact.security":"Security|||Report a vulnerability without exposing exploit details publicly","trust-centre.transparency.position.eyebrow":"Public position","trust-centre.transparency.position.heading":"Voluntary participation|||without invented classification","trust-centre.transparency.position.body":"Dweve records Loom as a general-purpose AI model under Article 3(63). The Commission's 10²³-FLOP criterion is indicative, not a safe harbour. No public designation of Loom as a GPAI model with systemic risk has been found as of 1 August 2026.","trust-centre.transparency.position.pills":"Signatory|||Taskforce|||Voluntary full scope","trust-centre.transparency.position.chrome":"Status boundary|||Dweve public position|||Current record|||Precise, not overstated","trust-centre.position.signatory":"Voluntary signatory|||Dweve has chosen the Code path before external release","trust-centre.position.taskforce":"Taskforce participant|||Dweve contributes through the official signatory working forum","trust-centre.position.designation":"No public designation found|||No public Commission record currently states that Loom has been designated as a GPAI model with systemic risk","trust-centre.position.code":"Code implementation|||Public model, training, copyright, downstream, safety, and monitoring records","trust-centre.position.architecture":"Architecture-aware record|||Dweve records Loom as a constraint-learning cognitive graph, not as one floating-point transformer; the distinction does not remove the statutory classification analysis","trust-centre.position.scope":"Voluntary safety scope|||Dweve applies the full discipline while no public designation is recorded","trust-centre.transparency.chapters.eyebrow":"Implementation map","trust-centre.transparency.chapters.heading":"Three GPAI chapters|||one operational record","trust-centre.transparency.chapters.body":"Transparency covers capabilities and the downstream boundary. Copyright covers acquisition and rights controls. Safety and security cover evaluation, risk, and monitoring.","trust-centre.transparency.chapters.pills":"Transparency|||Copyright|||Safety and security","trust-centre.transparency.chapters.chrome":"Voluntary coverage|||Implementation matrix|||Public and internal|||All workstreams visible","trust-centre.chapter.transparency":"Transparency|||Model documentation, training summary, downstream information, and version history","trust-centre.chapter.copyright":"Copyright|||Source policy, robots controls, rights reservations, and a complaint route","trust-centre.chapter.safety":"Safety and security|||Framework, evaluations, risk acceptance, incidents, and post-release monitoring","trust-centre.chapter.model":"Model documentation|||Identity, architecture, modalities, limits, state, and access routes","trust-centre.chapter.data":"Training-content summary|||772 source families are public; admitted-item decisions, provenance and evidence are maintained in the controlled Spindle ledger","trust-centre.chapter.risk":"Safety framework|||Voluntary risk scenarios, evaluation gates, and acceptance records prepared","trust-centre.chapter.monitoring":"Post-release monitoring|||Usage signals, feedback, incidents, changes, and corrective action are release controls","trust-centre.transparency.governance.eyebrow":"Evidence lifecycle","trust-centre.transparency.governance.heading":"Observe, decide, publish|||then preserve the history","trust-centre.transparency.governance.body":"A public page is the final surface, not the control itself. Dweve records the observation, assessment, approval, evidence package, publication version, and later correction as separate states.","trust-centre.transparency.governance.pills":"Owner|||Evidence|||Version","trust-centre.transparency.governance.chrome":"Record ownership|||Publication workflow|||Four gates|||Every revision attributable","trust-centre.governance.owner":"Dweve owns the record|||The organisation is Dweve and the accountable public publisher is Dweve","trust-centre.governance.version":"Every page is versioned|||Public files and human-readable pages carry the same release number","trust-centre.governance.evidence":"Claims need evidence|||A status can remain prepared until the backing artefact is verified","trust-centre.governance.observe":"Observe|||Capture a model, source, usage, evaluation, incident, or policy change","trust-centre.governance.assess":"Assess|||Determine scope, evidence, affected records, and release consequence","trust-centre.governance.approve":"Approve|||Assign an accountable owner and accept or reject the public statement","trust-centre.governance.publish":"Publish|||Update HTML, JSON, feeds, and retained history together","trust-centre.transparency.boundaries.eyebrow":"Claim discipline","trust-centre.transparency.boundaries.heading":"Evidence records what happened|||without asking readers to assume","trust-centre.transparency.boundaries.body":"A trace proves the recorded route. A signature proves identity and integrity. Deterministic replay proves equality under its captured contract. None of those facts alone proves that a use was lawful, fair, safe, or beneficial.","trust-centre.transparency.boundaries.pills":"Scope|||Evidence|||Limits","trust-centre.transparency.boundaries.chrome":"Interpretation guide|||What records can prove|||Three boundaries|||No blanket compliance claim","trust-centre.boundary.claims":"Scoped claims|||Every statement names the model, route, version, or control it describes","trust-centre.boundary.placeholders":"Publication discipline|||Unknown values are never guessed; future events are dated and their method and evidence contract stay visible","trust-centre.boundary.history":"Corrections remain visible|||A replacement does not erase the previous public statement","trust-centre.boundary.proves":"Record evidence|||Identity, integrity, lineage, configuration, execution, and captured replay","trust-centre.boundary.supports":"Assessment support|||Controls can support legal, safety, procurement, and audit assessment","trust-centre.boundary.never":"No automatic conclusion|||Evidence does not turn every deployment or decision into a compliant one","trust-centre.models.registry.eyebrow":"Registry entry","trust-centre.models.registry.heading":"The model family is singular|||the record is not vague","trust-centre.models.registry.body":"Dweve has one model family in this public Dweve register: Dweve Loom 1.0. Products such as Fabric, Aura, Mesh, and Spindle are systems around it, not additional foundation models.","trust-centre.models.registry.pills":"Provider|||Version|||Status","trust-centre.models.registry.chrome":"Registry identity|||Current public entry|||One model|||Adaptive state explained","trust-centre.model.provider":"Provider|||Dweve B.V., Arnhem, the Netherlands","trust-centre.model.version":"Version|||Dweve Loom 1.0","trust-centre.model.hash":"Adaptive state|||Loom 1.0 changes during use; an exact captured state receives its own complete manifest","trust-centre.model.name":"Dweve Loom 1.0|||Constraint-learning intelligence across cognitive threads and 528 top-level domain specialists","trust-centre.model.status":"Current availability|||Internal-only pre-release testing","trust-centre.model.release":"Planned external availability|||Invite-only from 1 September 2026","trust-centre.model.designation":"Regulatory designation|||No public Commission designation found for a GPAI model with systemic risk (current record)","trust-centre.models.access.eyebrow":"Access surface","trust-centre.models.access.heading":"One model can enter|||through several governed routes","trust-centre.models.access.body":"Invitees can use Loom individually or in combinations through Fabric, Aura, API, licensed local deployment, and Mesh. Each route changes the surrounding responsibility and evidence package, not the model identity.","trust-centre.models.access.pills":"Hosted|||API|||Licensed","trust-centre.models.access.chrome":"Release route|||Planned access map|||Five surfaces|||One model identity","trust-centre.access.invite":"Invite-only|||External participation is selected, not publicly open","trust-centre.access.external":"Planned 1 September 2026|||External testing begins with testers, partners, investors, and invited participants","trust-centre.access.local":"Licensed local operation|||Eligible deployments receive a licensed package and captured starting-state manifest","trust-centre.access.fabric":"Fabric|||The web workspace and human operating surface","trust-centre.access.aura":"Aura|||The closed-loop development runtime around chosen models and tools","trust-centre.access.api":"API|||A direct programmatic model and system boundary","trust-centre.access.licensed":"Licensed deployment|||Local, private, or isolated operation under the selected licence","trust-centre.access.mesh":"Mesh|||Distributed execution and deployment within the Dweve architecture","trust-centre.models.release.eyebrow":"Release readiness","trust-centre.models.release.heading":"Availability follows evidence|||not the other way around","trust-centre.models.release.body":"The external access date does not convert incomplete fields into claims. Artefact identity, training-source details, evaluation records, public controls, and operating runbooks must reach their release state independently.","trust-centre.models.release.pills":"Identity|||Sources|||Evaluation","trust-centre.models.release.chrome":"Readiness controls|||Model release readiness|||Four stages|||Evidence records the status","trust-centre.gate.hash":"Captured-state manifest|||Required for a reproducible evaluation, investigation, export, or replay bundle","trust-centre.gate.sources":"Training-source register|||772 source-family entries published; exact item, version, rights and provenance evidence is controlled in Spindle","trust-centre.gate.evals":"Evaluation package|||Prepared for release-specific results and acceptance evidence","trust-centre.gate.internal":"Internal operation|||Current pre-release system remains within Dweve","trust-centre.gate.complete":"Complete evidence|||Run the versioned route checks and bind actual pass evidence before external opening","trust-centre.gate.publish":"Publish invite release|||Open the approved routes and freeze the matching public record","trust-centre.gate.monitor":"Monitor operation|||Record feedback, usage, incidents, evaluations, and material changes","trust-centre.loom.experts":"Top-level domain-specialist catalogue|||528 domain specialists, with deeper specialist components, solvers, perception models, and other components","trust-centre.loom.selection":"Observed routing average|||Around 4 to 8 top-level domain specialists per request, never a hard minimum or maximum","trust-centre.loom.state":"Adaptive during use|||Loom continuously gains governed constraints; captured states identify what a particular run used","trust-centre.loom.anatomy.eyebrow":"Cognitive weave","trust-centre.loom.anatomy.heading":"The task selects the graph|||not one permanent model path","trust-centre.loom.anatomy.body":"Constraint learning supplies the governed domain state. Loom then connects only the typed threads a request actually needs and records the graph that ran.","trust-centre.loom.anatomy.pills":"Sparse routing|||Typed state|||Recorded path","trust-centre.loom.anatomy.chrome":"How Loom composes|||Cognitive execution graph|||Task-shaped|||More than the specialist count","trust-centre.loom.catalogue":"528 top-level domains|||The number describes the domain-specialist catalogue, not every cognitive component in Loom","trust-centre.loom.average":"Around 4 to 8 selected|||An observed average that varies with the question, state, and available resources","trust-centre.loom.components":"Nested capabilities|||Specialists can use sub-specialists, perception, retrieval, memory, solvers, and verification","trust-centre.loom.perception":"Perception|||Turns raw multimodal input into typed spans, regions, events, entities, fields, and signals","trust-centre.loom.memory":"Memory|||Supplies bounded working, episodic, semantic, and procedural state","trust-centre.loom.expert":"Domain specialist|||Applies specialist knowledge and constraints relevant to the task","trust-centre.loom.solver":"Reasoning and solvers|||Runs the problem form that fits the task instead of forcing everything into prediction","trust-centre.loom.verifier":"Verification|||Checks evidence, constraints, results, and the route before emission","trust-centre.loom.modalities.eyebrow":"Modality boundary","trust-centre.loom.modalities.heading":"Loom interprets broadly|||the outputs stay specific","trust-centre.loom.modalities.body":"The public boundary is asymmetric. Loom may interpret video, but it does not generate it. The documented output scope is text, code, and the images Dweve ships.","trust-centre.loom.modalities.pills":"Interpret broadly|||Documented outputs|||No video generation","trust-centre.loom.modalities.chrome":"Input and output|||Modality matrix|||Asymmetric boundary|||Video input is supported","trust-centre.modality.input":"Interpretation|||All represented modalities can be interpreted through the appropriate perception route","trust-centre.modality.output":"Generation|||The documented release scope is text, code, and the images Dweve ships","trust-centre.modality.video":"Video boundary|||Video can be interpreted, but Loom does not generate video","trust-centre.modality.language":"Natural language|||Interpretation and generation across textual language","trust-centre.modality.code":"Code|||Interpretation, structured reasoning, transformation, and generation","trust-centre.modality.structured":"Structured data|||Typed structures can appear in supported envelopes; no separate output route is claimed","trust-centre.modality.image":"Images|||Dweve ships image output with C2PA Content Credentials","trust-centre.modality.audio":"Audio|||Audio may be interpreted where documented; no generic audio-output route is claimed","trust-centre.modality.sensor":"Sensor data|||Signals and events may be interpreted where documented; no separate output route is claimed","trust-centre.modality.videoInput":"Video input|||Temporal visual interpretation without video generation","trust-centre.loom.state.eyebrow":"Memory and adaptation","trust-centre.loom.state.heading":"Loom changes during use|||the record captures the state","trust-centre.loom.state.body":"Deterministic does not mean frozen. Loom can consolidate, decay, compact, forget, and adapt runtime state. The sealed model and policy artefacts are identified separately from the memory snapshot and execution state used for a captured run.","trust-centre.loom.state.pills":"Adaptive|||Versioned|||Capturable","trust-centre.loom.state.chrome":"State boundary|||Four memory responsibilities|||Policy-bounded|||A snapshot is part of replay","trust-centre.state.changes":"Adaptive runtime|||Loom's memory and task state can change as it is used","trust-centre.state.sealed":"Captured state|||Specialist catalogue, constraints, sources, policies, memory, and executable package are bound when exact reconstruction is required","trust-centre.state.capture":"Captured contract|||Exact replay includes the applicable runtime state and memory snapshot","trust-centre.state.working":"Working memory|||The active task, intermediate values, and current execution context","trust-centre.state.episodic":"Episodic memory|||Prior cases and outcomes retained under the applicable policy","trust-centre.state.semantic":"Semantic memory|||Durable facts and relationships available to the relevant thread","trust-centre.state.procedural":"Procedural memory|||Reusable skills, methods, and ways of working","trust-centre.loom.replay.eyebrow":"Deterministic contract","trust-centre.loom.replay.heading":"Same sealed input and state|||same bytes across architectures","trust-centre.loom.replay.body":"Dweve produces byte-for-byte identical results across supported architectures. The complete state manifest pins the graph, routing, ordering, seeds, versions, memory state, constraints, solvers, evidence, and package integrity.","trust-centre.loom.replay.pills":"Complete manifest|||Captured state|||Byte-identical","trust-centre.loom.replay.chrome":"Replay manifest|||What must be pinned|||Seven fields|||Live world changes stay explicit","trust-centre.replay.artifact":"Captured-state identity|||Assigned to the exact specialist, constraint, source, configuration, memory, tool, solver, and policy manifest","trust-centre.replay.architecture":"Cross-architecture result|||Byte-for-byte identical under the supported captured deterministic contract","trust-centre.replay.live":"Live replay boundary|||A new run may consult changed external evidence or changed runtime state unless those are pinned","trust-centre.replay.graph":"Execution graph|||The exact threads and typed edges used by the task","trust-centre.replay.routing":"Routing and order|||Selected specialists, route scores, tie-breaking, and execution order","trust-centre.replay.seed":"Seeds and versions|||Explicit randomness, model versions, policies, and configuration","trust-centre.replay.memory":"Memory snapshot|||The runtime state against which the captured run executed","trust-centre.replay.evidence":"Evidence and tools|||Sources, tool results, constraints, solvers, and verification results","trust-centre.loom.operation.eyebrow":"Data during use","trust-centre.loom.operation.heading":"User data is not raw training|||explicit questions, declared route","trust-centre.loom.operation.body":"Prompts, conversations, files, tool inputs, and private workspace context are not raw training material and are not silently mined into constraints. When Loom cannot answer from its governed state it asks an explicit question and tells the user that the answer will train the governed constraint system. Only a derived candidate may enter Spindle.","trust-centre.loom.operation.pills":"No raw user-data training|||Question route disclosed|||Feedback guides research","trust-centre.loom.operation.chrome":"Operational data boundary|||From request to governed discovery|||Four separations|||Private context stays out","trust-centre.operation.retention":"No silent user-data mining|||Prompts, conversations, uploads, documents, tool inputs, and private context are not raw training material or silently mined into constraints; an explicitly asked-for answer can yield only a derived candidate through Spindle","trust-centre.operation.feedback":"Feedback guides direction|||Ordinary optional feedback can guide future mining but is not training data or a constraint; the explicit question route is separate","trust-centre.operation.local":"Local deterministic parity|||Licensed local artefacts follow the same captured deterministic contract","trust-centre.operation.sourceBoundary":"External-source boundary|||Only an independently fetched source enters ordinary candidate review; an explicitly asked-for answer follows the disclosed derived-candidate route","trust-centre.operation.inputs":"User input|||Processed for the requested task; an explicit knowledge-gap question says the answer will train the governed constraint system and only a derived candidate can enter Spindle","trust-centre.operation.tools":"External source discovery|||Winnow can discover a source during a search; candidate review is disclosed in the same response","trust-centre.operation.trace":"Execution trace|||Records what the weave actually consulted without publishing private chain-of-thought","trust-centre.operation.feedbackPath":"Ordinary feedback path|||Feedback still needs an admissible external source, extraction, validation, evidence, and Spindle promotion; it is not itself a constraint","trust-centre.operation.questionRoute":"Explicit question route|||When Loom cannot answer from governed state, it tells the user an answer will train the governed constraint system; only a validated derived constraint may be stored","trust-centre.training.template":"Public template|||All expected source, period, processing, rights, and governance fields are present","trust-centre.training.list":"Source catalogue|||772 supplied entries across 19 source groups","trust-centre.training.release":"Evidence boundary|||The public catalogue explains the source families and admission rules; exact item, version, rights and provenance records remain controlled in Spindle","trust-centre.training.chain.eyebrow":"Training evidence route","trust-centre.training.chain.heading":"Acquisition is one step|||promotion is another","trust-centre.training.chain.body":"Winnow gathers candidate material under declared source rules. Spindle preserves source identity, rights, and lineage, then governs what may be promoted. Loom consumes only the versioned result.","trust-centre.training.chain.pills":"Candidate|||Evidence|||Promotion","trust-centre.training.chain.chrome":"Product responsibility|||Winnow to Spindle to Loom|||Three stages|||Collection is not automatic inclusion","trust-centre.training.policy":"Source policy first|||Collection begins only after the applicable access and operator rules are available","trust-centre.training.provenance":"Evidence remains attached|||Address, time, status, content identity, and processing state follow the material","trust-centre.training.promotion":"Promotion is governed|||Candidate material does not become training material merely because it was collected","trust-centre.training.winnow":"Winnow|||Discovers, fetches, parses, and records candidate material under shared controls","trust-centre.training.spindle":"Spindle|||Maintains provenance, quality, policy, disagreement, and promotion state","trust-centre.training.loom":"Loom training workflow|||Consumes only the versioned material and policy state selected for that release","trust-centre.training.register.eyebrow":"Source-family register","trust-centre.training.register.heading":"The catalogue is public|||admission stays item-specific","trust-centre.training.register.body":"Dweve has supplied 772 source-family entries across 19 groups. The public catalogue names them and states each licence gate, usable slice, and purpose. A catalogue entry is not a claim that every release or item was used by Loom 1.0.","trust-centre.training.register.pills":"772 entries|||19 groups|||Evidence-bound","trust-centre.training.register.chrome":"Register status|||Supplied catalogue and release boundary|||Five admission classes|||UNKNOWN means QUARANTINE","trust-centre.training.status":"Register status|||CURRENT SOURCE CATALOGUE: ITEM EVIDENCE HELD IN SPINDLE","trust-centre.training.noInvent":"No blanket admission|||A named source family enters no release until its exact item passes every critical gate","trust-centre.training.owner":"Publication owner|||Dweve verifies and publishes the final register","trust-centre.training.sourceNames":"Source names and families|||772 supplied entries published in the source inventory","trust-centre.training.categories":"Content categories and proportions|||772 source-family entries across 19 groups: A0 184 (23.8%), A1 125 (16.2%), A2 57 (7.4%), F 372 (48.2%) and R 34 (4.4%). These are catalogue proportions, not model weights.","trust-centre.training.periods":"Collection and creation periods|||Governed collection runs from 2023 to the present. Underlying works range from verified public-domain history to current sources; Spindle records the date and version of each admitted item.","trust-centre.training.rights":"Rights basis and reservation handling|||A0, A1, A2, F and R gates cover public domain, attribution, reciprocal terms, record-level filtering and named rights instruments. robots.txt, access terms, reservations, objections and withdrawal state are enforced and recorded.","trust-centre.training.processing":"Filtering, deduplication, transformation, and enrichment|||Winnow acquires, Reed parses, BitWeave retrieves, Spindle and Lattice govern, Ledger and AION preserve and verify evidence, FMI and Twin test state and behaviour, and Selvedge controls executable material. Repeated corroboration precedes constraint promotion.","trust-centre.training.personal":"Personal-data presence and safeguards|||User prompts, conversations, files, documents, tool input and private workspace context are not raw training material or silently mined into constraints. An explicitly asked-for answer follows the disclosed derived-candidate route; candidate sources must pass privacy, confidentiality and minimisation gates.","trust-centre.training.languages":"Languages, regions, and modality distribution|||Multilingual and global, with strong Dutch, English and European official-source coverage. Text, code, structured records, formal material, images, audio, sensors, science, geospatial data and simulations are admitted by purpose and provenance.","trust-centre.training.optOut":"Exclusion, objection, and removal effects|||A valid reservation, rights failure, privacy issue or rightsholder decision blocks future acquisition, quarantines affected candidates and routes promoted constraints for review, correction or withdrawal through their Spindle lineage.","trust-centre.training.controls.eyebrow":"Acquisition controls","trust-centre.training.controls.heading":"The route records the decision|||at every hand-off","trust-centre.training.controls.body":"Winnow enforces robots.txt with deny-by-default behavior, per-domain rate limits, and per-request audit logging. Operator lawful basis and retention stay separate decisions.","trust-centre.training.controls.pills":"Robots|||Audit|||Promotion gate","trust-centre.training.controls.chrome":"Control ownership|||Candidate processing line|||Five gates|||Operator decisions stay explicit","trust-centre.training.robots":"Robots before fetch|||Uncached domains are refused until robots.txt has been fetched and parsed","trust-centre.training.audit":"Per-request record|||Source, URL, timestamp, status, and content hash are recorded","trust-centre.training.decision":"Operator responsibility|||Lawful basis, minimisation, purpose, and retention are not decided by the crawler","trust-centre.training.discover":"Discover|||Find candidate sources through the configured engines, catalogues, feeds, APIs, or files","trust-centre.training.fetch":"Fetch|||Apply source policy, rate, target-safety, and request controls","trust-centre.training.parse":"Parse|||Turn the response into typed records while preserving source identity","trust-centre.training.assess":"Assess|||Spindle evaluates provenance, evidence quality, policy, conflicts, and suitability","trust-centre.training.promote":"Promote|||A versioned decision admits material into the governed training state","trust-centre.training.publication.eyebrow":"Publication readiness","trust-centre.training.publication.heading":"The catalogue is public|||the ledger stays controlled","trust-centre.training.publication.body":"The human- and machine-readable source catalogue is public. Exact item, licence, attribution, withdrawal and source-version evidence remains in Spindle for audit and controlled disclosure; it is not duplicated as a public release ledger.","trust-centre.training.publication.pills":"Catalogue public|||Controls applied|||History preserved","trust-centre.training.publication.chrome":"Disclosure boundary|||Public summary and controlled evidence|||Two layers|||Useful without exposing every internal record","trust-centre.training.hash":"Covered model identity|||Dweve Loom 1.0; exact source admission is bound through release and captured-state manifests","trust-centre.training.summary":"Summary record|||Published in canonical HTML and machine-readable JSON","trust-centre.training.history":"Revision history|||Every material source-register change will enter the Trust change feed","trust-centre.training.prepare":"Publish the summary|||Name source families, modalities, periods, purposes, rights controls and processing in human- and machine-readable form","trust-centre.training.receive":"Keep item evidence controlled|||Bind exact source, version, treatment, attribution and withdrawal state in Spindle for audit and authorised access","trust-centre.training.verify":"Verify before promotion|||Corroborate candidate facts, resolve contradictions and require domain or human review where the evidence demands it","trust-centre.training.publish":"Maintain one current record|||Keep content version 1.0 until launch; record material public changes in the Trust ledger thereafter","trust-centre.copyright.robots":"Robots policy|||Deny by default until robots.txt is fetched and parsed","trust-centre.copyright.rates":"Per-domain rates|||Token buckets follow each source's documented policy","trust-centre.copyright.contact":"Rights channel|||A public reporting route accepts reservations, objections, and complaints","trust-centre.copyright.acquire.eyebrow":"Built-in posture","trust-centre.copyright.acquire.heading":"Politeness is enforced|||before content is returned","trust-centre.copyright.acquire.body":"The Winnow product record describes the controls that ship in the collection route. They are crawler controls, not a declaration that every operator purpose or later use is lawful.","trust-centre.copyright.acquire.pills":"Deny by default|||Rate-limited|||Audit-logged","trust-centre.copyright.acquire.chrome":"Crawler controls|||What happens before output|||Five controls|||Lawful basis remains separate","trust-centre.copyright.default":"Default posture|||A source without a cached robots decision is refused until the policy is retrieved","trust-centre.copyright.operator":"Operator decision|||The operator remains responsible for lawful basis, minimisation, purpose, and retention","trust-centre.copyright.log":"Request evidence|||Every request keeps its source, address, time, status, and content identity","trust-centre.copyright.ssrf":"Target safety|||Private and internal network ranges are blocked in URL input","trust-centre.copyright.lawful":"Lawful access only|||Winnow does not bypass passwords, paywalls, CAPTCHAs, or effective access restrictions","trust-centre.copyright.infringingSources":"Known infringement sources|||Applicable EU and EEA court or authority lists block persistently infringing commercial sources","trust-centre.copyright.pseudonym":"Personal-data handling|||Identifiers can be pseudonymised at the point of capture","trust-centre.copyright.audit":"Audit record|||The request route and result are exportable for review","trust-centre.copyright.roles.eyebrow":"Product roles","trust-centre.copyright.roles.heading":"Winnow collects|||Spindle decides what may persist","trust-centre.copyright.roles.body":"Loom does not gain an undocumented independent scraping route. It can invoke Winnow as a governed tool. Training workflows use Spindle to preserve provenance and control promotion into governed material.","trust-centre.copyright.roles.pills":"Acquire|||Govern|||Consume","trust-centre.copyright.roles.chrome":"Responsibility split|||Request-to-use chain|||Four hand-offs|||No secret crawler","trust-centre.copyright.winnowRole":"Winnow owns acquisition|||Source policy, fetching, parsing, target safety, rates, and request records","trust-centre.copyright.spindleRole":"Spindle owns governance|||Provenance, quality, policy, conflict, withdrawal, and promotion state","trust-centre.copyright.loomRole":"Loom owns task use|||It consumes governed state or calls Winnow through a recorded tool boundary","trust-centre.copyright.request":"Task requests evidence|||A user or workflow asks for current or collected information","trust-centre.copyright.collect":"Winnow executes|||The request passes the applicable source and network controls","trust-centre.copyright.govern":"Spindle evaluates|||The material remains candidate state until policy permits promotion","trust-centre.copyright.consume":"Loom uses the result|||The trace records the source or tool result actually consulted","trust-centre.copyright.outputGuard":"Protected-source reproduction|||The rights suite tests reproduction, source-rights bypass, refusal, provenance, and traceability before a route opens","trust-centre.copyright.outputRule":"Acceptable-use boundary|||Infringing use, evading reservations, and unlawful reproduction are prohibited","trust-centre.copyright.rightsholder.eyebrow":"Rightsholder channel","trust-centre.copyright.rightsholder.heading":"A concern needs a receipt|||and a reviewable outcome","trust-centre.copyright.rightsholder.body":"Rightsholders can identify the work, source, reservation, crawler event, requested action, and supporting evidence. The reporting desk records the submission and routes the matter without requiring sensitive evidence in the first message.","trust-centre.copyright.rightsholder.pills":"Identify|||Investigate|||Respond","trust-centre.copyright.rightsholder.chrome":"Complaint workflow|||Rights request route|||Four stages|||Evidence can follow securely","trust-centre.copyright.channel":"Public route|||The Trust reporting desk accepts copyright and source-access matters","trust-centre.copyright.receipt":"Submission reference|||A successful form submission returns a recorded identifier","trust-centre.copyright.outcome":"Reasoned response|||The review records scope, findings, action, and any effect on governed material","trust-centre.copyright.submit":"Submit|||Identify the work, source, reservation, concern, and requested outcome","trust-centre.copyright.triage":"Triage|||Separate access policy, source identity, rights, removal, and security issues","trust-centre.copyright.investigate":"Investigate|||Trace the request, source record, Spindle state, downstream use, and relevant version","trust-centre.copyright.respond":"Respond|||State the finding, action, limits, and route for further evidence or challenge","trust-centre.copyright.machine.eyebrow":"Machine-readable policy","trust-centre.copyright.machine.heading":"Humans and systems should read|||the same current rule","trust-centre.copyright.machine.body":"The crawler policy is published as JSON and linked to the Trust change feed. Winnow identifies itself publicly, follows robots.txt and source-specific rights controls, and does not bypass passwords, paywalls or CAPTCHAs.","trust-centre.copyright.machine.pills":"JSON|||Atom|||Versioned","trust-centre.copyright.machine.chrome":"Public endpoints|||Crawler information record|||Three formats|||Named user agent","trust-centre.copyright.userAgent":"Winnow public user agent|||Dweve-Winnow/1.0 (+https://dweve.com/trust/copyright)","trust-centre.copyright.feed":"Change notifications|||The Atom feed carries material policy and record updates","trust-centre.copyright.policyVersion":"Current policy version|||1.0, dated 1 August 2026","trust-centre.copyright.machineJson":"Crawler-policy JSON|||Machine-readable controls, roles, user-agent identity and reporting route","trust-centre.copyright.changeFeed":"Trust Atom feed|||A subscribable record of material public changes","trust-centre.copyright.humanPolicy":"Human-readable policy|||This page explains the same controls and boundaries in context","trust-centre.downstream.package":"Information package|||Model identity, capabilities, limitations, interfaces, evaluation context, and changes","trust-centre.downstream.contact":"Request channel|||A deployer can ask for missing or integration-specific information","trust-centre.downstream.boundary":"Responsibility boundary|||Model evidence does not decide the deployer's purpose or operating controls","trust-centre.downstream.package.eyebrow":"Common package","trust-centre.downstream.package.heading":"Publish once|||scope only what is private","trust-centre.downstream.package.body":"The public record should answer the common questions before a private exchange begins. Integration secrets, customer configuration, and sensitive evaluation evidence can then be scoped to the requesting downstream provider.","trust-centre.downstream.package.pills":"Current|||Versioned|||Scoped","trust-centre.downstream.package.chrome":"Information surface|||Downstream document set|||Six records|||Public first","trust-centre.downstream.current":"Current package|||Records are tied to the model and Trust content version","trust-centre.downstream.versioned":"Change-aware|||Material changes enter the public history and machine feed","trust-centre.downstream.scoped":"Need-to-know detail|||Sensitive integration information is scoped to a legitimate request","trust-centre.downstream.modelRecord":"Model record|||Identity, architecture, modalities, state, replay, and operation","trust-centre.downstream.capabilities":"Capabilities|||What Loom can interpret, generate, compose, and verify","trust-centre.downstream.limitations":"Limitations|||No video generation, state and live-evidence boundaries, and use constraints","trust-centre.downstream.integration":"Integration information|||Interfaces, input and output contracts, tool boundaries, and deployment posture","trust-centre.downstream.evaluation":"Evaluation context|||Release-specific cases, methods, results, limits, and acceptance state","trust-centre.downstream.changes":"Change information|||What changed, why, who is affected, and which record replaced it","trust-centre.downstream.request.eyebrow":"Information request","trust-centre.downstream.request.heading":"Ask for the missing boundary|||not another brochure","trust-centre.downstream.request.body":"A request should identify the integration, decision, model version, needed field, and reason. Dweve then answers from the public record, provides scoped material, or explains why not.","trust-centre.downstream.request.pills":"Check|||Scope|||Answer","trust-centre.downstream.request.chrome":"Request handling|||Public-to-private route|||Four stages|||Every answer is attributable","trust-centre.downstream.publicFirst":"Check public records first|||The common package is designed to reduce repeated private requests","trust-centre.downstream.private":"Sensitive details stay scoped|||Security, customer, and proprietary information needs a controlled exchange","trust-centre.downstream.auditTrail":"Request and answer are recorded|||The information version and stated boundary remain traceable","trust-centre.downstream.check":"Check|||Identify the public record and exact missing field","trust-centre.downstream.ask":"Request|||Describe the downstream system, purpose, version, and information need","trust-centre.downstream.scope":"Scope|||Confirm authority, confidentiality, affected interfaces, and evidence level","trust-centre.downstream.answer":"Answer|||Provide the public record, a scoped evidence package, the factual lifecycle state, or a reasoned confidentiality limit","trust-centre.downstream.responsibility.eyebrow":"Responsibility map","trust-centre.downstream.responsibility.heading":"Dweve owns the model boundary|||deployers own the use boundary","trust-centre.downstream.responsibility.body":"Responsibility is not transferred by a PDF. Dweve owns accurate model and service information. Deployers own purpose, data, people, configuration, oversight, and local operation. Shared interfaces and incidents require both sides.","trust-centre.downstream.responsibility.pills":"Provider|||Deployer|||Shared","trust-centre.downstream.responsibility.chrome":"Operating boundary|||Who owns which decision|||Three zones|||Evidence follows ownership","trust-centre.downstream.dweve":"Dweve boundary|||Model version, captured-state contracts, provider documentation, service controls, provider changes, and provider incidents","trust-centre.downstream.deployer":"Deployer boundary|||Purpose, input data, roles, configuration, local policy, human oversight, and affected people","trust-centre.downstream.shared":"Shared boundary|||Integration contracts, evaluation, monitoring, incident exchange, and corrective action","trust-centre.downstream.modelBoundary":"Model and service|||The record identifies what Loom and the selected Dweve route do and do not do","trust-centre.downstream.useBoundary":"Purpose and decision|||The deployer decides whether the use is legitimate, proportionate, and appropriately overseen","trust-centre.downstream.operationBoundary":"Production operation|||Each deployment names its owners for identity, data, security, change, recovery, evidence, and incidents","trust-centre.generated.trace":"Execution trace contract|||Route, component, evidence, tool, policy and state fields bind each output to its actual execution contract","trust-centre.generated.marking":"Output marking|||The visible and machine-readable contract is public; each route must pass it before external opening","trust-centre.generated.video":"Video generation|||Not supported","trust-centre.generated.outputs.eyebrow":"Output surface","trust-centre.generated.outputs.heading":"Documented outputs are narrow|||input interpretation is broader","trust-centre.generated.outputs.body":"Input and output are not the same capability. Loom interprets video but never generates it. The release scope is text, code, and shipped images.","trust-centre.generated.outputs.pills":"Input|||Output|||Exclusion","trust-centre.generated.outputs.chrome":"Capability boundary|||Generated-output matrix|||Three documented output classes|||No generated video","trust-centre.generated.interpret":"Interpretation scope|||The represented input boundary includes video","trust-centre.generated.generate":"Generation scope|||Text, code, and the images Dweve ships","trust-centre.generated.noVideo":"Explicit exclusion|||Loom does not generate video","trust-centre.generated.language":"Natural language|||Answers, explanations, summaries, transformations, and structured prose","trust-centre.generated.code":"Code|||Generated or transformed code within the selected tool and policy boundary","trust-centre.generated.structured":"Structured data|||Not named in the current release output scope","trust-centre.generated.image":"Images|||Generated visual output with its applicable disclosure route","trust-centre.generated.audio":"Audio|||No generic Dweve audio-output route is claimed","trust-centre.generated.sensor":"Sensor-derived output|||Not named in the current release output scope","trust-centre.generated.markers.eyebrow":"Disclosure route","trust-centre.generated.markers.heading":"Mark the surface|||preserve the deeper trace","trust-centre.generated.markers.body":"A visible disclosure informs the recipient, a machine-readable marker preserves the origin, and the route trace provides the deeper record. The contract is public now; the actual product version and export paths are tested before a route opens externally.","trust-centre.generated.markers.pills":"Visible|||Machine-readable|||Traceable","trust-centre.generated.markers.chrome":"Output anatomy|||Creation-to-verification route|||Four layers|||Marking does not replace provenance","trust-centre.generated.visible":"Visible disclosure|||Every applicable user surface makes the AI interaction or generated nature clear where the recipient receives it","trust-centre.generated.machine":"Machine-readable disclosure|||Every supported response and export carries origin and trace-reference information in metadata, its envelope, or an inseparable sidecar","trust-centre.generated.create":"Create|||Loom produces or transforms output under the selected execution contract","trust-centre.generated.attach":"Attach trace|||Bind model, graph, state, evidence, tools, policy and product-route fields to the output record","trust-centre.generated.label":"Disclose|||Present the human-readable context and preserve machine-readable origin in every supported output route","trust-centre.generated.verify":"Verify|||Round-trip marker, trace, exports and captured replay against Loom 1.0 and the actual product version","trust-centre.generated.disclosure.eyebrow":"Deployment responsibility","trust-centre.generated.disclosure.heading":"The marker must survive|||the route to the final user","trust-centre.generated.disclosure.body":"Dweve verifies disclosure fields on every Dweve-operated surface before external opening. Downstream deployers preserve or present them in the final context. Local operators control presentation, not the meaning of the origin record. The contract below is current even though no external route has opened yet.","trust-centre.generated.disclosure.pills":"Fabric|||API|||Local","trust-centre.generated.disclosure.chrome":"Who presents what|||Disclosure responsibility|||Three surfaces|||Origin must not disappear","trust-centre.generated.fabric":"Fabric, Aura, and Mesh|||Must pass interaction disclosure, visible context, marking persistence and trace access before external opening","trust-centre.generated.api":"API surface|||Must pass structured origin fields, trace references, format persistence and documentation before external access","trust-centre.generated.local":"Licensed local surface|||Must preserve equivalent disclosure and the byte-identical captured-record contract","trust-centre.generated.dweveDuty":"Dweve|||Create accurate fields, document them, and preserve them through Dweve-operated surfaces","trust-centre.generated.deployerDuty":"Downstream deployer|||Carry the disclosure into the actual user experience and decision context","trust-centre.generated.userDuty":"User or operator|||Do not remove, misstate, or detach the marker from the output it describes","trust-centre.use.allowed":"Legitimate governed work|||Use with a defined purpose, authority, owner, and review path","trust-centre.use.blocked":"Prohibited or uncontrolled work|||Requests that violate rights, law, safety, or authority are outside the boundary","trust-centre.use.enforcement":"Policy and review|||Technical holds, refusals, records, and human review form the response path","trust-centre.use.boundaries.eyebrow":"Permission before capability","trust-centre.use.boundaries.heading":"The system can help|||only inside an authorised purpose","trust-centre.use.boundaries.body":"Dweve supports research, analysis, creation, engineering, operations, and governed automation. The accountable user or organisation still has to define the legitimate purpose, data authority, affected people, review, and stop conditions.","trust-centre.use.boundaries.pills":"Purpose|||Authority|||Review","trust-centre.use.boundaries.chrome":"Use decision|||Allowed and prohibited boundary|||Two sides|||Capability is not permission","trust-centre.use.purpose":"Named purpose|||The work has an explicit objective and accountable owner","trust-centre.use.authority":"Authority to act|||Inputs, tools, systems, people, and decisions are within the user's authority","trust-centre.use.evidence":"Reviewable record|||Material decisions preserve evidence, uncertainty, human action, and outcome","trust-centre.use.allowedWork":"Within the boundary|||Legitimate assistance, analysis, creation, engineering, simulation, research, and controlled automation","trust-centre.use.prohibitedWork":"Outside the boundary|||Illegal, harmful, deceptive, rights-abusing, unauthorised, or uncontrolled high-impact action","trust-centre.use.prohibited.eyebrow":"Prohibited uses","trust-centre.use.prohibited.heading":"Some requests should stop|||before a tool is called","trust-centre.use.prohibited.body":"This public policy states the classes of use Dweve does not permit. Product and deployment controls still need the more specific rules, roles, thresholds, and review procedures that apply to each implementation.","trust-centre.use.prohibited.pills":"Rights|||Safety|||Authority","trust-centre.use.prohibited.chrome":"Public boundary|||Prohibited-use classes|||Six classes|||Specific controls follow deployment","trust-centre.use.noBypass":"No control bypass|||Do not defeat policy, security, disclosure, trace, or human-review controls","trust-centre.use.noHarm":"No deliberate harm|||Do not use Dweve to facilitate violence, exploitation, abuse, or dangerous wrongdoing","trust-centre.use.noRightsAbuse":"No rights abuse|||Do not unlawfully infringe privacy, copyright, equality, or other protected rights","trust-centre.use.illegal":"Illegal activity|||Planning, enabling, concealing, or scaling conduct prohibited by applicable law","trust-centre.use.malware":"Malicious cyber activity|||Credential theft, unauthorised access, destructive code, evasion, or exploitation","trust-centre.use.deception":"Deception and impersonation|||Fraud, manipulative impersonation, undisclosed synthetic identity, or falsified evidence","trust-centre.use.surveillance":"Unlawful surveillance|||Tracking, profiling, or inferring people without the required authority and safeguards","trust-centre.use.infringement":"Copyright infringement|||Using acquisition or generation routes to evade rights reservations or unlawfully reproduce protected work","trust-centre.use.unsafeAutomation":"Uncontrolled high-impact action|||Delegating consequential decisions or physical actions without appropriate authority, checks, and human responsibility","trust-centre.use.response.eyebrow":"Control response","trust-centre.use.response.heading":"Refusal is one state|||review and appeal complete it","trust-centre.use.response.body":"A policy concern can end in a refusal, a hold, a human review, a narrower route, or a confirmed block.","trust-centre.use.response.pills":"Hold|||Review|||Resolve","trust-centre.use.response.chrome":"Enforcement path|||From signal to outcome|||Four stages|||Human challenge remains possible","trust-centre.use.refuse":"Refuse or hold|||The system does not continue through the questioned capability boundary","trust-centre.use.record":"Record the control|||The policy, request class, state, and resulting action remain attributable","trust-centre.use.appeal":"Challenge route|||A user can provide missing authority, context, or evidence through the reporting desk","trust-centre.use.detect":"Detect|||A policy rule, tool boundary, user report, or operator review identifies a concern","trust-centre.use.hold":"Hold|||Prevent action while preserving enough context for a safe review","trust-centre.use.review":"Review|||Assess purpose, authority, evidence, potential harm, and a narrower alternative","trust-centre.use.resolve":"Resolve|||Allow, narrow, refuse, suspend, correct, or escalate with a recorded reason","trust-centre.report.channel":"Single entry point|||All Trust topics use one accessible submission surface","trust-centre.report.receipt":"Recorded submission|||The contact system returns a submission reference after acceptance","trust-centre.report.routing":"Topic ownership|||The selected concern determines the internal review route","trust-centre.report.routes.eyebrow":"Choose the subject","trust-centre.report.routes.heading":"One form|||six accountable routes","trust-centre.report.routes.body":"Start with the category that best matches the issue. Dweve can re-route it internally without asking you to resubmit. Describe sensitive evidence first and wait for a controlled exchange before sending secrets or exploit details.","trust-centre.report.routes.pills":"Copyright|||Safety|||Information","trust-centre.report.routes.chrome":"Public intake|||Trust reporting routes|||Six topics|||One recorded hand-off","trust-centre.report.public":"Accessible intake|||No account is required to open the reporting desk","trust-centre.report.sensitive":"Sensitive evidence|||Do not place secrets, personal data, or exploit instructions in the first message","trust-centre.report.followup":"Controlled follow-up|||Dweve can request the evidence through an appropriate route after triage","trust-centre.report.copyright":"Copyright and source access|||Rights reservations, crawler conduct, attribution, removal, or protected content","trust-centre.report.downstream":"Downstream information|||Missing model, capability, limitation, evaluation, or integration information","trust-centre.report.safety":"Model safety|||Unexpected capability, harmful behavior, failed control, or post-release risk signal","trust-centre.report.security":"Security vulnerability|||A flaw in software, access, isolation, integrity, or a tool boundary","trust-centre.report.privacy":"Privacy concern|||Personal data, retention, access, correction, deletion, or deployment privacy","trust-centre.report.model":"Model quality|||Incorrect, inconsistent, unsupported, or misleading model behavior","trust-centre.report.form.eyebrow":"Recorded intake","trust-centre.report.form.lead":"Describe the concern","trust-centre.report.form.accent":"without losing its route","trust-centre.report.form.body":"Choose one topic and give enough context to identify the affected page, model version, source, output, integration, or event. You can provide sensitive supporting material after Dweve confirms the appropriate channel.","trust-centre.report.form.pill.1":"One form","trust-centre.report.form.pill.2":"Topic routed","trust-centre.report.form.pill.3":"Reference returned","trust-centre.report.form.side":"Submission contract","trust-centre.report.form.receipt":"Reference|||Successful delivery returns a recorded submission identifier","trust-centre.report.form.evidence":"Evidence later|||The first message should describe sensitive supporting material, not expose it","trust-centre.report.form.reply":"Follow-up|||Dweve requests the next evidence or clarifies the outcome through the supplied address","trust-centre.report.form.footer":"The first message opens the record; it does not need to contain the entire case","trust-centre.report.form.topic":"Concern type","trust-centre.report.form.name":"Name","trust-centre.report.form.email":"Email","trust-centre.report.form.message":"What happened, where, and what should Dweve inspect?","trust-centre.report.form.submit":"Submit the record","trust-centre.report.form.sending":"Sending","trust-centre.report.form.sent":"Received. Your submission reference was created by the contact system.","trust-centre.report.form.error":"Check your name, email, and message, then try again.","trust-centre.report.form.privacy":"Do not include passwords, private keys, personal records, or exploit instructions in this first message.","trust-centre.report.form.routeTitle":"Selected route","trust-centre.report.form.record":"Record","trust-centre.report.form.recordValue":"Contact submission","trust-centre.report.form.status":"Initial state","trust-centre.report.form.statusValue":"Awaiting triage","trust-centre.report.form.id":"Reference","trust-centre.report.topic.copyright":"Copyright","trust-centre.report.topic.copyright.body":"Sources, rights reservations, crawler events, attribution, removal, or alleged infringement.","trust-centre.report.topic.downstream":"Downstream information","trust-centre.report.topic.downstream.body":"Model, capability, limitation, interface, evaluation, or change information for an integration.","trust-centre.report.topic.safety":"Model safety","trust-centre.report.topic.safety.body":"Unexpected capability, harmful behavior, control failure, misuse pattern, or post-release risk signal.","trust-centre.report.topic.security":"Security","trust-centre.report.topic.security.body":"A vulnerability or weakness in software, access, isolation, integrity, deployment, or tool boundaries.","trust-centre.report.topic.privacy":"Privacy","trust-centre.report.topic.privacy.body":"A question or concern about personal data, retention, access, correction, deletion, or privacy controls.","trust-centre.report.topic.model":"Model quality","trust-centre.report.topic.model.body":"Incorrect, inconsistent, unsupported, misleading, or otherwise unexpected Loom behavior.","trust-centre.changes.version":"Trust content version|||1.0","trust-centre.changes.feed":"Machine feeds|||JSON and Atom records are published beside the human page","trust-centre.changes.history":"Retained history|||Material revisions do not silently replace their predecessor","trust-centre.changes.current.eyebrow":"Current release","trust-centre.changes.current.heading":"Version 1.0 establishes|||the public Trust record","trust-centre.changes.current.body":"The first release creates the route structure, Loom model record, prepared training summary, crawler and copyright record, downstream package, disclosure boundary, acceptable-use policy, reporting desk, and machine-readable endpoints.","trust-centre.changes.current.pills":"1 August 2026|||Trust 1.0|||Pre-release","trust-centre.changes.current.chrome":"Release facts|||Version 1.0 change set|||Four entries|||No breaking model change","trust-centre.changes.date":"Publication date|||1 August 2026","trust-centre.changes.scope":"Affected surface|||New Trust Centre and linked machine-readable records","trust-centre.changes.breaking":"Model behavior|||No model-behavior change is asserted by this content release","trust-centre.changes.hub":"Trust Centre created|||Public index, status vocabulary, evidence chain, release line, and reporting route","trust-centre.changes.records":"Loom records prepared|||Model, modalities, state, replay, operation, training template, and downstream information","trust-centre.changes.machine":"Machine endpoints added|||AI transparency, model, training, crawler, and change records in JSON plus Atom","trust-centre.changes.pending":"Publication state|||No literal holding text remains; future events are dated facts and exact item evidence stays controlled in Spindle","trust-centre.changes.versioning.eyebrow":"Version model","trust-centre.changes.versioning.heading":"Content and model versions|||are not the same thing","trust-centre.changes.versioning.body":"Trust content version 1.0 identifies this pre-launch disclosure. Dweve Loom 1.0 identifies the evolving model. Captured-state manifests identify exact evaluation, investigation, export, or replay states. None substitutes for another.","trust-centre.changes.versioning.pills":"Content|||Model|||Artefact","trust-centre.changes.versioning.chrome":"Identity layers|||What a version means|||Four fields|||Exact reconstruction uses a captured-state manifest","trust-centre.changes.contentVersion":"Trust content version|||The public document and machine-record release, currently 1.0","trust-centre.changes.modelVersion":"Model version|||Dweve Loom 1.0","trust-centre.changes.artifactHash":"Captured-state identity|||Recorded where a change, evaluation, investigation, export, or replay requires an exact state","trust-centre.changes.what":"What changed|||The specific page, field, control, model, source policy, or machine endpoint","trust-centre.changes.why":"Why it changed|||Correction, evidence update, release, control change, incident, or clearer boundary","trust-centre.changes.effect":"Who is affected|||Users, deployers, evaluators, rightsholders, or internal operators","trust-centre.changes.archive":"What it replaced|||The prior public version remains identifiable and retrievable","trust-centre.changes.feeds.eyebrow":"Subscribe and compare","trust-centre.changes.feeds.heading":"One change|||three readable forms","trust-centre.changes.feeds.body":"The HTML page is for people, JSON is for systems and comparisons, and Atom is for notification. They share the same release identity and point back to the canonical Trust record.","trust-centre.changes.feeds.pills":"HTML|||JSON|||Atom","trust-centre.changes.feeds.chrome":"Publication formats|||Change endpoints|||Three surfaces|||One version","trust-centre.changes.human":"Human page|||Context, explanation, links, and visual status","trust-centre.changes.json":"JSON record|||Structured versions, entries, statuses, affected routes and factual lifecycle states","trust-centre.changes.atom":"Atom feed|||A subscribable notification for material Trust updates","trust-centre.changes.page":"Trust change page|||The canonical human-readable history","trust-centre.changes.jsonLink":"Change JSON|||The current machine-readable ledger","trust-centre.changes.atomLink":"Change Atom feed|||The machine-readable subscription endpoint","trust-centre.machine.label":"Machine records","trust-centre.machine.transparency":"AI transparency JSON","trust-centre.machine.model":"Loom model JSON","trust-centre.machine.training":"Training register JSON","trust-centre.machine.sources":"772-source inventory JSON","trust-centre.machine.admission":"Admission policy JSON","trust-centre.machine.crawler":"Crawler policy JSON","trust-centre.machine.changes":"Change ledger JSON","trust-centre.machine.assurance":"Assurance index JSON","trust-centre.machine.downstream":"Downstream pack JSON","trust-centre.machine.output":"Output transparency JSON","trust-centre.machine.marking":"Output marking policy JSON","trust-centre.machine.complianceCost":"Compliance cost disclosure JSON","trust-centre.machine.releaseGate":"Release-readiness JSON","trust-centre.machine.integrity":"SHA-256 manifest","trust-centre.page.safety.crumb":"Safety and security","trust-centre.page.safety.hero.eyebrow":"Safety and security framework","trust-centre.page.safety.hero.title":"A control is only useful|||when its evidence can be inspected","trust-centre.page.safety.hero.lede":"Dweve applies the full safety and security discipline voluntarily. This record separates model, tool, deployment, and operator controls, then shows which operating evidence is ready and which case records still have to be published.","trust-centre.page.safety.hero.actions":"Inspect the control layers|||Report a security concern","trust-centre.page.safety.glyph":"Safety control register|||Pre-release","trust-centre.page.evaluations.crumb":"Evaluations","trust-centre.page.evaluations.hero.eyebrow":"Evaluation and testing record","trust-centre.page.evaluations.hero.title":"Results need an identity|||before they become a claim","trust-centre.page.evaluations.hero.lede":"Every evaluation identifies Loom, the exact suite, configuration, captured state, evidence and reviewer decision. The complete method and acceptance model are public now. No first external-release result exists because that release has not occurred.","trust-centre.page.evaluations.hero.actions":"Inspect the evaluation schema|||Read the Loom record","trust-centre.page.evaluations.glyph":"Evaluation register|||Current method","trust-centre.page.risk.crumb":"Risk management","trust-centre.page.risk.hero.eyebrow":"Model and operational risk","trust-centre.page.risk.hero.title":"Risk is a decision record|||not a colour on a matrix","trust-centre.page.risk.hero.lede":"The risk record connects a scenario to evidence, treatment, ownership, residual conditions, and reassessment. No public Commission designation of Loom as a GPAI model with systemic risk has been found as of the current record date; Dweve still applies the complete voluntary discipline without using architecture as an escape clause.","trust-centre.page.risk.hero.actions":"Inspect the risk method|||Open the safety framework","trust-centre.page.risk.glyph":"Risk decision register|||Current framework","trust-centre.page.monitoring.crumb":"Post-market monitoring","trust-centre.page.monitoring.hero.eyebrow":"Monitoring and model use","trust-centre.page.monitoring.hero.title":"Release opens the record|||it does not close it","trust-centre.page.monitoring.hero.lede":"Loom is currently in internal-only testing. This record prepares the signals, confidentiality boundaries, adaptive-state snapshots, and review route needed for invite-only access and continued operation from 1 September 2026.","trust-centre.page.monitoring.hero.actions":"Inspect the monitoring loop|||Open the incident record","trust-centre.page.monitoring.glyph":"Monitoring register|||Pre-release","trust-centre.page.incidents.crumb":"Incidents","trust-centre.page.incidents.hero.eyebrow":"Incident response and publication","trust-centre.page.incidents.hero.title":"One report starts a case|||not a public conclusion","trust-centre.page.incidents.hero.lede":"Safety, security, privacy, copyright, model-quality and downstream concerns enter one route. Evidence is preserved before a case is classified. No public incident entry has been published as of 1 August 2026; that is a publication status, not a zero-incident claim.","trust-centre.page.incidents.hero.actions":"Inspect the incident route|||Report a concern","trust-centre.page.incidents.glyph":"Incident register|||Current · no public entry","trust-centre.page.data.crumb":"Data governance","trust-centre.page.data.hero.eyebrow":"Data and operating boundaries","trust-centre.page.data.hero.title":"User data is not raw training|||explicit questions are disclosed","trust-centre.page.data.hero.lede":"Normal tester interactions are not stored or used by Dweve by default. If Loom cannot answer from governed state, it asks an explicit knowledge-gap question and tells the user that the answer will train the governed constraint system; only a derived candidate can enter Spindle.","trust-centre.page.data.hero.actions":"Inspect the data boundaries|||Open the privacy route","trust-centre.page.data.glyph":"Data boundary register|||Current policy","trust-centre.page.documentation.crumb":"Technical documentation","trust-centre.page.documentation.hero.eyebrow":"Technical documentation index","trust-centre.page.documentation.hero.title":"One release identity|||across every evidence layer","trust-centre.page.documentation.hero.lede":"The documentation index binds Loom 1.0 to its model, training, evaluation, risk, safety, monitoring, incident, downstream and environmental records. Known facts, bounded estimates and genuine future events are kept distinct.","trust-centre.page.documentation.hero.actions":"Inspect the evidence index|||Read downstream information","trust-centre.page.documentation.glyph":"Technical evidence index|||Version 1.0","trust-centre.hub.assurance.eyebrow":"Operational assurance","trust-centre.hub.assurance.heading":"The model card is the start|||not the whole trust record","trust-centre.hub.assurance.body":"A serious public record also shows how Dweve evaluates, accepts risk, monitors use, responds to incidents, governs data, and controls technical documentation. These records expose their pre-release state instead of presenting prepared processes as completed evidence.","trust-centre.hub.assurance.pills":"Assess|||Operate|||Improve","trust-centre.hub.assurance.chrome":"Assurance scope|||Operational evidence map|||Seven records|||Prepared is not proven","trust-centre.assurance.public":"Public structure|||Every assurance workstream has a stable route and visible status","trust-centre.assurance.operational":"Operational evidence|||Case records and measured results are published only after verification","trust-centre.assurance.placeholders":"Publication complete|||Methods, boundaries and schemas are public; results appear only from actual versioned events","trust-centre.assurance.safety":"Safety and security|||Control layers, lifecycle, threat areas, and disclosure route","trust-centre.assurance.evaluations":"Evaluations|||Versioned test identity, families, replay, review, and results","trust-centre.assurance.risk":"Risk management|||Scenario, evidence, treatment, acceptance, and reassessment","trust-centre.assurance.monitoring":"Post-market monitoring|||Signals, protected usage information, adaptive state, and review","trust-centre.assurance.incidents":"Incidents|||Intake, preservation, triage, notification, and public history","trust-centre.assurance.data":"Data governance|||Interaction, feedback, training, trace, state, and local boundaries","trust-centre.assurance.documentation":"Technical documentation|||Release identity and access layers across every dossier","trust-centre.safety.framework":"Framework status|||Control structure prepared for the planned external release","trust-centre.safety.designation":"Voluntary scope|||No public Commission systemic-risk designation found; complete discipline applied by choice","trust-centre.safety.channel":"Disclosure channel|||Security and safety concerns have a public intake route","trust-centre.safety.scope.eyebrow":"Control perimeter","trust-centre.safety.scope.heading":"Four control layers|||four different owners","trust-centre.safety.scope.body":"Model behaviour, tool access, deployment configuration, and operator authority cannot be collapsed into one safety score. The public record separates what Dweve controls directly from what an integrator or operator must configure and govern.","trust-centre.safety.scope.pills":"Model|||Tools|||Operation","trust-centre.safety.scope.chrome":"Control ownership|||Safety perimeter|||Four layers|||Boundaries stay explicit","trust-centre.safety.modelLayer":"Model layer|||Architecture, capability limits, refusal, state, trace, and replay controls","trust-centre.safety.toolLayer":"Tool layer|||Declared tools, permissions, source policy, results, and execution evidence","trust-centre.safety.deploymentLayer":"Deployment layer|||Identity, configuration, isolation, access, monitoring, and update policy","trust-centre.safety.modelControl":"Model control|||Dweve owns released Loom artefacts and their documented behaviour","trust-centre.safety.toolControl":"Tool control|||Dweve products mediate tool calls through recorded policy boundaries","trust-centre.safety.deploymentControl":"Deployment control|||The selected operator owns environment-specific access and configuration","trust-centre.safety.operatorControl":"Operator control|||Purpose, authority, review, and action remain human and organisational duties","trust-centre.safety.lifecycle.eyebrow":"Control lifecycle","trust-centre.safety.lifecycle.heading":"A mitigation moves|||only with its evidence","trust-centre.safety.lifecycle.body":"A control begins as a stated risk treatment and becomes releasable only after its implementation, verification, owner, version, and remaining boundary are recorded. Later evidence can reopen the decision.","trust-centre.safety.lifecycle.pills":"Identify|||Mitigate|||Verify","trust-centre.safety.lifecycle.chrome":"Control governance|||Evidence-bearing lifecycle|||Five gates|||No silent acceptance","trust-centre.safety.owner":"Control owner|||Named ownership field is prepared per release and case","trust-centre.safety.evidence":"Evidence attachment|||Tests, traces, decisions, and implementation records travel together","trust-centre.safety.versioning":"Version boundary|||A control result identifies the model, state, configuration, and date","trust-centre.safety.identify":"Identify|||State the scenario, affected surface, and evidence needed","trust-centre.safety.assess":"Assess|||Separate observed facts, assumptions, severity, and uncertainty","trust-centre.safety.mitigate":"Mitigate|||Bind a technical or operational control to the stated scenario","trust-centre.safety.verify":"Verify|||Run the defined check against the identified release contract","trust-centre.safety.publish":"Publish|||Expose the result, boundary, owner, and next review at the right access level","trust-centre.safety.threats.eyebrow":"Threat register structure","trust-centre.safety.threats.heading":"Threat families organise the work|||evidence decides the outcome","trust-centre.safety.threats.body":"The register groups work without claiming that a category has passed. Every actual finding needs its own evidence, affected version, treatment, and status.","trust-centre.safety.threats.pills":"Misuse|||Integrity|||Resilience","trust-centre.safety.threats.chrome":"Prepared taxonomy|||Threat workboard|||Six families|||No invented pass marks","trust-centre.safety.noScores":"No invented scores|||A documented method is never presented as a test result","trust-centre.safety.noInvent":"No invented coverage|||A short public list never claims to cover Loom's full capability surface","trust-centre.safety.caseEvidence":"Case evidence|||Every actual test or incident receives its own versioned evidence bundle; no fictional case record is shown","trust-centre.safety.misuse":"Misuse and harmful capability|||Deliberate abuse, prohibited purpose, and control bypass scenarios","trust-centre.safety.toolAccess":"Tool and action access|||Permission, scope, target, side effect, and confirmation boundaries","trust-centre.safety.dataExposure":"Data exposure|||Sensitive input, state, output, trace, and tool-result handling","trust-centre.safety.integrity":"Integrity|||Artefact identity, configuration, evidence, update, and supply-path tampering","trust-centre.safety.availability":"Availability and control|||Resource pressure, degraded operation, recovery, and safe stopping","trust-centre.safety.supplyChain":"Supply chain|||Dependencies, build provenance, distribution, deployment, and update trust","trust-centre.safety.disclosure.eyebrow":"Coordinated disclosure","trust-centre.safety.disclosure.heading":"Preserve the finding first|||publish only what is safe","trust-centre.safety.disclosure.body":"The public form records the route without asking a reporter to publish exploit details. Dweve can request protected evidence, scope the affected system, and separate remediation communication from later public disclosure.","trust-centre.safety.disclosure.pills":"Report|||Preserve|||Respond","trust-centre.safety.disclosure.chrome":"Disclosure route|||Security case flow|||Four stages|||Sensitive evidence stays protected","trust-centre.safety.publicRoute":"Public intake|||The reporting desk is available without a private invitation","trust-centre.safety.sensitiveRoute":"Protected follow-up|||Sensitive details move only after Dweve establishes the case route","trust-centre.safety.outcome":"Outcome record|||Status and publishable conclusions follow verification","trust-centre.safety.submit":"Submit|||Describe the affected route, version, behaviour, and safe reproduction context","trust-centre.safety.preserve":"Preserve|||Retain evidence and avoid destructive or public reproduction","trust-centre.safety.triage":"Triage|||Confirm scope, severity, ownership, and immediate containment needs","trust-centre.safety.respond":"Respond|||Remediate, notify where required, verify, and publish the appropriate record","trust-centre.evaluations.schema":"Evaluation method|||Identity, configuration, state, evidence, acceptance and review fields are current","trust-centre.evaluations.results":"First external-release result|||No external-release run has occurred as of 1 August 2026","trust-centre.evaluations.identity.eyebrow":"Test identity","trust-centre.evaluations.identity.heading":"A score without a run contract|||is not a result","trust-centre.evaluations.identity.body":"Loom changes during use, so an evaluation pins the complete captured-state and runtime contract. The public record separates the evolving model identity from the evaluated capture. Suite and run identifiers are created only for an actual execution.","trust-centre.evaluations.identity.pills":"Artefact|||State|||Evidence","trust-centre.evaluations.identity.chrome":"Result identity|||Evaluation manifest|||Six fields|||Replay before comparison","trust-centre.evaluations.replay":"Captured replay|||A repeated captured contract must produce byte-identical results on supported architectures","trust-centre.evaluations.version":"Suite version|||Every claim identifies the exact evaluation definition","trust-centre.evaluations.hash":"Model identity|||Dweve Loom 1.0, adaptive during use","trust-centre.evaluations.model":"Model version|||Dweve Loom 1.0","trust-centre.evaluations.artifact":"Evaluated capture|||Time-bound state manifest assigned per evaluation","trust-centre.evaluations.suite":"Evaluation suite|||Every actual run receives a unique suite name and frozen version; no dummy identifier is published","trust-centre.evaluations.configuration":"Configuration and policy|||Graph, specialists, solvers, constraints, evidence, routing, tools, memory, policy, hardware and deployment posture","trust-centre.evaluations.state":"Captured state|||Snapshot identity and included memory classes","trust-centre.evaluations.evidence":"Evidence bundle|||Inputs, outputs, traces, tool results, comparisons, and reviewer decision","trust-centre.evaluations.families.eyebrow":"Evaluation coverage","trust-centre.evaluations.families.heading":"Test families show the coverage|||they do not manufacture a pass","trust-centre.evaluations.families.body":"Each suite freezes its cases and thresholds before execution. A critical failure in one family cannot be hidden by a stronger score elsewhere.","trust-centre.evaluations.families.pills":"Capability|||Safety|||Robustness","trust-centre.evaluations.families.chrome":"Prepared coverage|||Evaluation families|||Six workstreams|||Results remain separate","trust-centre.evaluations.categories":"Coverage structure|||The categories organise evidence across the model and deployment boundary","trust-centre.evaluations.thresholds":"Acceptance thresholds|||Frozen per versioned suite; critical safety, security, rights, authority, trace or replay failures block the affected route","trust-centre.evaluations.claims":"Claim boundary|||A prepared category is not evidence that Loom passed it","trust-centre.evaluations.capability":"Capability and task performance|||What Loom can interpret, generate, combine, and complete under a defined contract","trust-centre.evaluations.limitations":"Limitations and refusal|||Unsupported, uncertain, prohibited, and escalation behaviour","trust-centre.evaluations.bias":"Bias and rights impact|||Differential outcomes, representation, accessibility, and review context","trust-centre.evaluations.robustness":"Robustness and adversarial behaviour|||Perturbation, malformed input, conflicting evidence, and control pressure","trust-centre.evaluations.security":"Security|||Tool abuse, data exposure, integrity, access, supply chain, and deployment controls","trust-centre.evaluations.trace":"Trace and explanation|||Whether the recorded route, evidence, state, and result support inspection","trust-centre.evaluations.reproducibility.eyebrow":"Reproducibility boundary","trust-centre.evaluations.reproducibility.heading":"Replay the captured run|||that is not a changed world","trust-centre.evaluations.reproducibility.body":"A fully captured run replays byte for byte. A live rerun can differ.","trust-centre.evaluations.reproducibility.pills":"Pin|||Replay|||Compare","trust-centre.evaluations.reproducibility.chrome":"Replay contract|||Reproducibility route|||Four steps|||Live change stays visible","trust-centre.evaluations.captured":"Captured contract|||All result-bearing inputs and state must be identified","trust-centre.evaluations.live":"Live boundary|||Changed evidence or adaptive state creates a different run contract","trust-centre.evaluations.crossArchitecture":"Supported architectures|||The same captured local contract is byte-identical across supported architectures","trust-centre.evaluations.pin":"Pin|||Identify the release artefact, graph, state, policy, tools, and evidence","trust-centre.evaluations.run":"Run|||Execute the versioned suite without substituting an unrecorded input","trust-centre.evaluations.compare":"Compare|||Distinguish byte equality, semantic assessment, and acceptance decision","trust-centre.evaluations.explain":"Explain|||Publish the traceable reason for the result and its remaining boundary","trust-centre.evaluations.publication.eyebrow":"Result publication","trust-centre.evaluations.publication.heading":"The method is public now|||results belong to real runs","trust-centre.evaluations.publication.body":"Dweve does not prefill a result with a marketing estimate. Public summaries identify the matching model, captured state, method, decision and limitations. Sensitive test material can remain protected while its scope, severity, conclusion and release effect are recorded.","trust-centre.evaluations.publication.pills":"Run|||Review|||Release","trust-centre.evaluations.publication.chrome":"Publication readiness|||Evaluation release line|||Four states|||No estimate becomes a score","trust-centre.evaluations.publicResult":"Public result|||None issued as of 1 August 2026 because the first external release has not occurred","trust-centre.evaluations.sensitiveResult":"Protected evidence|||Detailed attack material can be restricted without hiding the tested scope","trust-centre.evaluations.history":"Version history|||Superseded methods and results remain identifiable","trust-centre.evaluations.prepare":"Prepare|||Freeze the suite definition, evidence requirements, and acceptance rule","trust-centre.evaluations.execute":"Execute|||Create the actual run identity and preserve inputs, expected outcomes, outputs, trace and tool evidence","trust-centre.evaluations.review":"Review|||Record the real reviewer, decision, severity, residual risk, affected routes and required retest","trust-centre.evaluations.release":"Release|||Publish the verified summary, method, boundary, and linked artefact","trust-centre.risk.register":"Risk register|||Current method, risk families, evidence rules, decision authority and reassessment triggers","trust-centre.risk.designation":"Regulatory position|||No public Commission systemic-risk designation found for Loom (current record)","trust-centre.risk.voluntary":"Voluntary discipline|||Dweve applies the complete scope by choice","trust-centre.risk.taxonomy.eyebrow":"Risk scope","trust-centre.risk.taxonomy.heading":"Start with the affected system|||not a generic severity label","trust-centre.risk.taxonomy.body":"The taxonomy separates model capability, misuse, rights, security, downstream integration, operation, and environmental evidence. A category organises review; it does not determine severity without a case.","trust-centre.risk.taxonomy.pills":"Scenario|||Evidence|||Impact","trust-centre.risk.taxonomy.chrome":"Risk perimeter|||Current taxonomy|||Seven families|||Case decides severity","trust-centre.risk.model":"Model risk|||Capability, limitation, adaptation, state, and evidence quality","trust-centre.risk.use":"Use risk|||Purpose, authority, affected people, human review, and downstream action","trust-centre.risk.operation":"Operational risk|||Deployment, access, availability, integrity, monitoring, and change","trust-centre.risk.capability":"Capability|||Unexpected combinations, inadequate limits, or misunderstood competence","trust-centre.risk.misuse":"Misuse|||Deliberate abuse, control bypass, prohibited purpose, or harmful automation","trust-centre.risk.rights":"Rights and society|||Privacy, copyright, discrimination, accessibility, and affected-party recourse","trust-centre.risk.security":"Safety and security|||Data, tool, artefact, supply-chain, infrastructure, and response risk","trust-centre.risk.downstream":"Downstream|||Integration, local configuration, operator authority, and reliance beyond scope","trust-centre.risk.environment":"Environment|||Current estate disclosed; historical training best estimate 18 MWh, plausible range 4 to 50 MWh, low confidence, with method and exclusions published","trust-centre.risk.process.eyebrow":"Decision lifecycle","trust-centre.risk.process.heading":"Treatment is not closure|||reassessment stays attached","trust-centre.risk.process.body":"A risk decision preserves the scenario, evidence, treatment, residual conditions, approver, and next review. New evaluation, use, incident, model, or state evidence can reopen it.","trust-centre.risk.process.pills":"Measure|||Treat|||Reassess","trust-centre.risk.process.chrome":"Risk governance|||Decision route|||Five gates|||Residual conditions remain","trust-centre.risk.owner":"Decision authority|||Dweve B.V. owns release authority; the accountable domain owner signs the evidence and the release owner records accept or stop","trust-centre.risk.evidence":"Evidence set|||Observed facts remain separate from assumptions and uncertainty","trust-centre.risk.decision":"Acceptance record|||Conditions, authority, expiry, and reassessment are required","trust-centre.risk.identify":"Identify|||Describe the scenario, affected system, people, and operating context","trust-centre.risk.measure":"Measure|||Use the relevant evidence and state what remains unknown","trust-centre.risk.treat":"Treat|||Reduce likelihood, impact, exposure, or permission through a bound control","trust-centre.risk.accept":"Accept or stop|||Record authority, residual conditions, and release effect","trust-centre.risk.monitor":"Monitor|||Reopen when a trigger, expiry, change, evaluation, or incident occurs","trust-centre.risk.position.eyebrow":"Regulatory boundary","trust-centre.risk.position.heading":"No designation|||no excuse to do less","trust-centre.risk.position.body":"No public Commission designation of Loom as a GPAI model with systemic risk has been found as of the current record date. Dweve does not claim that statutory thresholds or capability-based assessment disappear because Loom uses a different architecture. It documents architecture-native measurements and voluntarily performs the full assurance work.","trust-centre.risk.position.pills":"Precise status|||Full scope|||Open method","trust-centre.risk.position.chrome":"Public position|||Designation and discipline|||Three statements|||No threshold escape","trust-centre.risk.noDesignation":"No public designation found|||The dated public record contains no Commission designation of Loom as systemic risk","trust-centre.risk.noThresholdEscape":"No architecture escape|||Different compute terms do not make statutory analysis disappear","trust-centre.risk.fullDiscipline":"Complete voluntary scope|||Safety, security, evaluation, monitoring, and incident work proceed anyway","trust-centre.risk.legalStatus":"Legal status|||No public Commission systemic-risk designation found (current record)","trust-centre.risk.architectureTerms":"Architecture record|||Sparse cognitive graph described in measurable, native terms","trust-centre.risk.publicAssessment":"Public assessment|||Current framework applies to Loom 1.0 before external access and throughout operation","trust-centre.risk.highRiskPack":"Conditional deployer pack|||Instructions, oversight, logging, accuracy and conformity evidence prepared for any downstream high-risk classification","trust-centre.risk.decisions.eyebrow":"Decision contract","trust-centre.risk.decisions.heading":"Show what a decision must hold|||without inventing one","trust-centre.risk.decisions.body":"The first external model release has not occurred, so this page does not manufacture a public release approval. It publishes the binding fields, authority and triggers that every real assessment must carry.","trust-centre.risk.decisions.pills":"Owner|||Evidence|||Expiry","trust-centre.risk.decisions.chrome":"Decision contract|||Risk acceptance record|||Eight required fields|||No sample approval","trust-centre.risk.recordId":"Decision identifier|||Assigned when an assessment is opened; no dummy identifier is published for the not-yet-occurred first external release","trust-centre.risk.approver":"Decision authority|||Dweve B.V. holds release authority; the named domain owner and release owner are recorded in the actual assessment","trust-centre.risk.reviewDate":"Review timing|||Required before the planned 1 September 2026 invite-only external access, then on expiry or any earlier reassessment trigger","trust-centre.risk.context":"Scenario and scope|||Affected release, route, deployment, people, and operating assumptions","trust-centre.risk.evidenceSet":"Evidence and uncertainty|||Sources, evaluations, incidents, traces, gaps, and confidence","trust-centre.risk.residual":"Residual risk|||What remains after the selected controls","trust-centre.risk.conditions":"Acceptance conditions|||Authority, limits, monitoring, stop conditions, and expiry","trust-centre.risk.reassessment":"Reassessment triggers|||Release, state, use, evidence, incident, threshold, or date change","trust-centre.monitoring.internal":"Current phase|||Internal-only pre-release testing","trust-centre.monitoring.external":"Next phase|||Invite-only external access planned for 1 September 2026","trust-centre.monitoring.usage":"Usage information|||Content-free field register, aggregation, retention control and access matrix are public","trust-centre.monitoring.phases.eyebrow":"Operating phases","trust-centre.monitoring.phases.heading":"Monitoring begins before access|||and changes with the audience","trust-centre.monitoring.phases.body":"Internal testing establishes the first evidence and current release readiness. Invite-only access adds external use, feedback, and incident signals. Continued operation adds trend review, reassessment, and public change history.","trust-centre.monitoring.phases.pills":"Internal|||Invite-only|||Continuous","trust-centre.monitoring.phases.chrome":"Monitoring posture|||Operating timeline|||Four phases|||1 September stays explicit","trust-centre.monitoring.current":"Current operation|||Dweve teams use the pre-release system internally","trust-centre.monitoring.invite":"External access|||Invite-only access follows the stated records and release-readiness checks","trust-centre.monitoring.continuous":"Ongoing review|||Signals, decisions, incidents, changes, and reassessments remain versioned","trust-centre.monitoring.internalPhase":"Internal testing|||Observe behaviour, exercise controls, and close public-record gaps","trust-centre.monitoring.gatePhase":"Release readiness|||Bind model identity, documentation, evaluation, monitoring, and response ownership","trust-centre.monitoring.invitePhase":"Invite-only operation|||Testers, partners, investors, and other invitees enter through selected routes","trust-centre.monitoring.operatingPhase":"Continued operation|||Review protected signals and publish material Trust changes","trust-centre.monitoring.signals.eyebrow":"Signal register","trust-centre.monitoring.signals.heading":"A signal starts a review|||it does not decide the case","trust-centre.monitoring.signals.body":"Monitoring joins volunteered feedback, submitted incidents, security reports, versioned evaluations, protected use information, and adaptive-state evidence. Each source keeps its collection boundary and confidence.","trust-centre.monitoring.signals.pills":"Observe|||Compare|||Escalate","trust-centre.monitoring.signals.chrome":"Signal boundaries|||Monitoring evidence map|||Six sources|||Purpose stays attached","trust-centre.monitoring.disclosed":"Disclosed collection|||A signal is collected only through a stated product, feedback, or reporting route","trust-centre.monitoring.minimised":"Purpose minimisation|||Only fields needed for the stated review should enter the case","trust-centre.monitoring.versioned":"Version context|||Signals identify the model, route, state, date, and available evidence","trust-centre.monitoring.feedback":"Optional feedback|||Material submitted through the explicit feedback route for Dweve's use","trust-centre.monitoring.incident":"Incident reports|||Safety, privacy, copyright, quality, and downstream concerns","trust-centre.monitoring.security":"Security reports|||Protected vulnerability and control-failure submissions","trust-centre.monitoring.evaluation":"Evaluation evidence|||Versioned capability, limitation, robustness, rights, and security checks","trust-centre.monitoring.usageSignal":"Model-use information|||Release context, workload bands, control events and review signals; prompts, files and raw content are excluded","trust-centre.monitoring.stateSignal":"Adaptive-state evidence|||Captured state and change identity where a review depends on adaptation","trust-centre.monitoring.usageInfo.eyebrow":"Protected usage information","trust-centre.monitoring.usageInfo.heading":"Measure what matters|||without exposing who used it","trust-centre.monitoring.usageInfo.body":"Use information can matter to risk assessment while remaining commercially and personally sensitive. Dweve defines the purpose first, excludes prompt and document content, minimises fields, separates identities and logs every assessment use.","trust-centre.monitoring.usageInfo.pills":"Minimise|||Aggregate|||Protect","trust-centre.monitoring.usageInfo.chrome":"Confidentiality design|||Usage-information boundary|||Four controls|||Content stays outside","trust-centre.monitoring.competition":"Commercial sensitivity|||Usage patterns can reveal customers, strategy, demand, and product position","trust-centre.monitoring.privacy":"Personal and organisational privacy|||Identifiers and content need a specific necessity before collection","trust-centre.monitoring.systemic":"Risk relevance|||Aggregated use can inform exposure and reassessment without publishing raw business data","trust-centre.monitoring.collectBoundary":"Collection boundary|||No prompt, upload, raw output or private context; each collected signal needs a declared purpose, retention rule and authorised role","trust-centre.monitoring.aggregate":"Aggregate|||Prefer exposure bands and trends over raw identifiable events where they answer the question","trust-centre.monitoring.separate":"Separate|||Keep reporting identity, customer identity, content, and risk signal apart unless linkage is necessary","trust-centre.monitoring.review":"Review|||Record who used the information, for which assessment, with which result","trust-centre.monitoring.adaptation.eyebrow":"Adaptive-state monitoring","trust-centre.monitoring.adaptation.heading":"Loom changes during use|||the release identity stays","trust-centre.monitoring.adaptation.body":"Monitoring separates the evolving Loom 1.0 identity from a captured adaptive state.","trust-centre.monitoring.adaptation.pills":"Artefact|||State|||Change","trust-centre.monitoring.adaptation.chrome":"Change identity|||Adaptive monitoring loop|||Five steps|||Model and state stay separate","trust-centre.monitoring.modelIdentity":"Model identity|||Dweve Loom 1.0 plus a captured-state manifest where a review needs exact reconstruction","trust-centre.monitoring.stateIdentity":"State identity|||Captured working, episodic, semantic, and procedural memory context","trust-centre.monitoring.changeIdentity":"Change record|||Policy, state, evidence, configuration, or model change identified separately","trust-centre.monitoring.observe":"Observe|||Receive a versioned signal through a disclosed route","trust-centre.monitoring.snapshot":"Snapshot|||Capture the state and execution context needed to inspect it","trust-centre.monitoring.compare":"Compare|||Replay the captured contract and distinguish changed-world evidence","trust-centre.monitoring.escalate":"Escalate|||Open evaluation, risk, incident, or release review when a trigger is met","trust-centre.monitoring.record":"Record|||Publish the material change or retain the protected decision evidence","trust-centre.incidents.channel":"Reporting channel|||One public intake with subject-specific routing","trust-centre.incidents.register":"Incident register|||Structure and change history prepared","trust-centre.incidents.entries":"Register snapshot|||No public entry published as of 1 August 2026; not a zero-incident claim","trust-centre.incidents.scope.eyebrow":"Incident perimeter","trust-centre.incidents.scope.heading":"Route by affected duty|||not by who reports it","trust-centre.incidents.scope.body":"One event can affect model behaviour, security, privacy, copyright, downstream integration, or several at once. The reporting route records the initial subject without preventing a case from being re-scoped after evidence review.","trust-centre.incidents.scope.pills":"Receive|||Scope|||Preserve","trust-centre.incidents.scope.chrome":"Case routing|||Incident intake map|||Six routes|||One case can cross duties","trust-centre.incidents.modelScope":"Model scope|||Version, behaviour, state, trace, evidence, and capability context","trust-centre.incidents.deploymentScope":"Deployment scope|||Configuration, access, operator action, integration, and local environment","trust-centre.incidents.evidenceScope":"Evidence scope|||Preserve useful context without submitting secrets in the first public message","trust-centre.incidents.safety":"Safety|||Harmful behaviour, unexpected capability, failed refusal, or unsafe action path","trust-centre.incidents.security":"Security|||Vulnerability, unauthorised access, integrity, supply-chain, or tool-control failure","trust-centre.incidents.privacy":"Privacy|||Unexpected retention, disclosure, identification, access, or rights concern","trust-centre.incidents.copyright":"Copyright|||Source, reservation, attribution, removal, output, or rightsholder concern","trust-centre.incidents.quality":"Model quality|||Incorrect, inconsistent, unsupported, misleading, or materially degraded behaviour","trust-centre.incidents.downstream":"Downstream|||Missing documentation, integration boundary, capability, limitation, or update information","trust-centre.incidents.process.eyebrow":"Response lifecycle","trust-centre.incidents.process.heading":"Preserve before classification|||learn after containment","trust-centre.incidents.process.body":"Evidence is secured before the case is classified, and the outcome feeds back into risk and monitoring rather than stopping at containment.","trust-centre.incidents.process.pills":"Triage|||Contain|||Learn","trust-centre.incidents.process.chrome":"Incident governance|||Case lifecycle|||Six stages|||Notification follows evidence","trust-centre.incidents.preserveFirst":"Evidence first|||Do not destroy the context needed to identify the affected contract","trust-centre.incidents.caseOwner":"Case ownership|||Owner and severity fields are assigned after initial validation","trust-centre.incidents.regulatory":"Notification boundary|||Applicable authorities and parties depend on the verified case","trust-centre.incidents.receive":"Receive|||Create the contact record and acknowledge the safe reporting route","trust-centre.incidents.preserve":"Preserve|||Secure the available artefact, state, trace, configuration, and report evidence","trust-centre.incidents.triage":"Triage|||Validate scope, affected duties, urgency, and responsible owners","trust-centre.incidents.contain":"Contain and remediate|||Limit exposure, correct the control, and prepare verification","trust-centre.incidents.notify":"Notify|||Inform affected parties or authorities when the verified case requires it","trust-centre.incidents.learn":"Learn|||Re-evaluate, update controls, monitor recurrence, and publish the appropriate history","trust-centre.incidents.registerView.eyebrow":"Public incident register","trust-centre.incidents.registerView.heading":"The register is live|||empty is not a safety claim","trust-centre.incidents.registerView.body":"No fake incident identifiers or empty zero-incident claim appears here. A public entry is added only from a verified case and records what can be disclosed without exposing people, customers, or exploitable detail.","trust-centre.incidents.registerView.pills":"Identity|||Status|||Outcome","trust-centre.incidents.registerView.chrome":"Publication schema|||Incident ledger|||Five fields|||No entry is not zero incidents","trust-centre.incidents.publicStatus":"Register status|||Current; no public incident entry published as of 1 August 2026","trust-centre.incidents.confidentiality":"Disclosure boundary|||Personal, customer, security, and investigation details can remain protected","trust-centre.incidents.history":"Change history|||Material updates to a public entry remain versioned","trust-centre.incidents.publicId":"Public incident identifier|||Created from the validated case without exposing the private case reference","trust-centre.incidents.date":"Event and publication dates|||Recorded separately and corrected visibly when evidence changes","trust-centre.incidents.affected":"Affected boundary|||Model, captured state, product route, deployment and relevant versions","trust-centre.incidents.status":"Case and remediation status|||Containment, correction, notification assessment, verification and closure state","trust-centre.incidents.outcome":"Publishable outcome|||Cause, effect, action, remaining boundary and learning without personal or exploitable detail","trust-centre.data.runtime":"Normal interaction|||Not stored or used by Dweve by default; explicit questions are separately disclosed","trust-centre.data.feedback":"Optional feedback|||Separate disclosed form and explicit choice","trust-centre.data.training":"Training-content register|||772 source families are public; item-level rights, provenance, admission and withdrawal records remain controlled in Spindle","trust-centre.data.boundaries.eyebrow":"Four data routes","trust-centre.data.boundaries.heading":"Purpose separates the records|||before storage does","trust-centre.data.boundaries.body":"Interaction data serves the requested task. An answer to an explicit knowledge-gap question follows a disclosed derived-constraint route. Optional feedback serves a separate improvement route, and training content belongs to a versioned governance record.","trust-centre.data.boundaries.pills":"Interact|||Choose|||Govern","trust-centre.data.boundaries.chrome":"Purpose boundary|||Data-route split|||Four records|||No silent reuse","trust-centre.data.separation":"Separate purposes|||Task execution, explicit question-derived constraints, optional feedback, and training-content governance have different records","trust-centre.data.purpose":"Selected authority|||A deployment or operator can have duties outside Dweve's own storage boundary","trust-centre.data.record":"Evidence of choice|||Feedback use requires the separate disclosed submission route","trust-centre.data.interactionData":"Normal interaction|||Processed for the requested task; not stored or reused by Dweve by default","trust-centre.data.feedbackData":"Optional feedback|||Used by Dweve only when deliberately submitted through the disclosed form","trust-centre.data.trainingData":"Training content|||Public source catalogue and admission rules, with exact item, version, rights and provenance decisions maintained in Spindle","trust-centre.data.runtimePolicy.eyebrow":"Runtime data","trust-centre.data.runtimePolicy.heading":"The execution can be traced|||without becoming training","trust-centre.data.runtimePolicy.body":"Loom uses input, tools, trace evidence, and adaptive state while performing a task. Those runtime functions do not authorise Dweve to retain or reuse tester interactions. The one exception is the disclosed knowledge-gap question. Licensed local operation stays inside the operator's boundary.","trust-centre.data.runtimePolicy.pills":"Execute|||Trace|||Separate","trust-centre.data.runtimePolicy.chrome":"Operating boundary|||Runtime data map|||Five surfaces|||No silent training use","trust-centre.data.noStorage":"Dweve storage|||Tester interactions are not retained by default","trust-centre.data.noTrainingUse":"Dweve use|||Tester interactions are not used for training by default; an explicit question states the derived-constraint route","trust-centre.data.localBoundary":"Local deployment|||Licensed local operation retains the same deterministic artefact contract within its own boundary","trust-centre.data.input":"Task input|||Used to perform the selected request","trust-centre.data.toolResult":"Tool result|||Used under the tool's declared permission, source, and execution boundary","trust-centre.data.trace":"Execution trace|||Records the route, evidence, state, tools, and decisions without publishing private chain-of-thought","trust-centre.data.adaptiveState":"Adaptive state|||Loom can change during use; exact state is captured only when evaluation, investigation, export, or replay requires it","trust-centre.data.localOperation":"Local operation|||Data and state remain inside the licensed operator's selected environment and policy boundary","trust-centre.data.feedbackRoute.eyebrow":"Optional feedback","trust-centre.data.feedbackRoute.heading":"A separate form for the choice|||visible and reversible by design","trust-centre.data.feedbackRoute.body":"Feedback is not required to use Loom. The route must say that Dweve will use the submitted material, describe the purpose, and keep that choice separate from normal task execution.","trust-centre.data.feedbackRoute.pills":"Optional|||Disclosed|||Recorded","trust-centre.data.feedbackRoute.chrome":"Consent boundary|||Feedback route|||Four steps|||No product-use condition","trust-centre.data.optional":"Not required|||Access to Loom does not depend on providing feedback","trust-centre.data.disclosed":"Dweve use stated|||The form explains that submitted material will be used by Dweve","trust-centre.data.revocation":"Withdrawal mechanics|||A feedback submitter can use the privacy route to stop future use and request removal of the identifiable feedback record, subject to legal or security preservation duties; independently discovered sources remain governed separately","trust-centre.data.choose":"Choose|||The tester deliberately opens the separate feedback route","trust-centre.data.disclose":"Disclose|||The form states purpose, use, required fields, and sensitive-data warning","trust-centre.data.submit":"Submit|||The feedback record captures the explicit action and supplied material","trust-centre.data.govern":"Govern|||Dweve applies the stated access, review, retention, and withdrawal record","trust-centre.data.rights.eyebrow":"Questions and rights","trust-centre.data.rights.heading":"One request route|||separate evidence behind it","trust-centre.data.rights.body":"A person can raise a privacy, feedback, source, or runtime-data concern through the reporting desk. Dweve verifies the relevant identity and scope before exposing, changing, or removing protected records.","trust-centre.data.rights.pills":"Request|||Verify|||Act","trust-centre.data.rights.chrome":"Rights route|||Data request flow|||Five steps|||Protected records stay protected","trust-centre.data.publicChannel":"Public intake|||The privacy topic routes the first message","trust-centre.data.identityCheck":"Protected verification|||Identity and authority depend on the requested action and record","trust-centre.data.actionRecord":"Outcome evidence|||The case records what was found, decided, changed, and communicated","trust-centre.data.request":"Request|||Describe the route, date, product, deployment, and desired outcome","trust-centre.data.locate":"Locate|||Identify which Dweve, local, feedback, source, or downstream boundary may hold the record","trust-centre.data.assess":"Assess|||Verify identity, authority, purpose, exceptions, and affected evidence","trust-centre.data.act":"Act|||Access, correct, restrict, remove, preserve, or explain as the verified case requires","trust-centre.data.confirm":"Confirm|||Return the outcome and retain the necessary case history","trust-centre.documentation.index":"Evidence index|||Public routes bind the technical record together","trust-centre.documentation.hash":"Captured-state manifest|||Assigned for evaluations, investigations, sealed packages, exports, and replay bundles","trust-centre.documentation.energy":"Energy record|||Historical training estimate published with method, range, exclusions and low-confidence boundary","trust-centre.documentation.records.eyebrow":"Evidence index","trust-centre.documentation.records.heading":"One model identity connects|||every technical dossier","trust-centre.documentation.records.body":"The index prevents a current model card from being paired with an old evaluation, source summary, or risk decision. Every dossier identifies its own version, status, access level, and matching Loom artefact.","trust-centre.documentation.records.pills":"Identity|||Evidence|||Access","trust-centre.documentation.records.chrome":"Documentation map|||Release evidence index|||Eight dossiers|||Cross-record versioning","trust-centre.documentation.public":"Public layer|||Stable human and machine routes expose the common record","trust-centre.documentation.downstream":"Downstream layer|||Integration-specific information is scoped and versioned for the receiving party","trust-centre.documentation.restricted":"Restricted layer|||Security, personal, customer, or confidential evidence can require controlled access","trust-centre.documentation.modelRecord":"Model record|||Architecture, modalities, adaptation, replay, operating data, and status","trust-centre.documentation.trainingRecord":"Training-content summary|||772 source families, five admission classes, deterministic admission rules and the controlled Spindle evidence boundary","trust-centre.documentation.evaluationRecord":"Evaluation record|||Suite identity, coverage, replay, results, and review","trust-centre.documentation.riskRecord":"Risk record|||Scenario, evidence, treatment, acceptance, and reassessment","trust-centre.documentation.safetyRecord":"Safety and security|||Control layers, threat work, lifecycle, and disclosure","trust-centre.documentation.monitoringRecord":"Monitoring|||Signals, usage-information boundary, adaptation, and review","trust-centre.documentation.incidentRecord":"Incidents|||Case route, response lifecycle, and public register","trust-centre.documentation.downstreamRecord":"Downstream information|||Capabilities, limitations, integration, evaluation, and changes","trust-centre.documentation.identity.eyebrow":"Release identity","trust-centre.documentation.identity.heading":"A version label is not enough|||the artefact must match","trust-centre.documentation.identity.body":"Dweve Loom 1.0 is the evolving model version. A time-bound captured-state manifest identifies the exact specialist, constraint, source, configuration, memory, tool, solver, and policy state needed for reconstruction. Trust content version 1.0 identifies this pre-launch public record.","trust-centre.documentation.identity.pills":"Model|||Artefact|||Record","trust-centre.documentation.identity.chrome":"Identity manifest|||Documentation release contract|||Six fields|||No inferred hash","trust-centre.documentation.contentVersion":"Trust content version|||1.0, updated 1 August 2026","trust-centre.documentation.modelVersion":"Model version|||Dweve Loom 1.0","trust-centre.documentation.modelHash":"State behaviour|||Adaptive during use; no single static hash identifies Loom as a whole","trust-centre.documentation.provider":"Provider|||Dweve B.V.","trust-centre.documentation.model":"Model|||Dweve Loom","trust-centre.documentation.version":"Model version|||1.0","trust-centre.documentation.hashValue":"Captured state|||A complete manifest is assigned whenever exact reconstruction is required","trust-centre.documentation.date":"Trust record date|||1 August 2026","trust-centre.documentation.compatibility":"Matching records|||Every evaluation, export, investigation and sealed deployment identifies its Loom version, captured state, configuration and applicable policy record","trust-centre.documentation.accessLayers.eyebrow":"Documentation access","trust-centre.documentation.accessLayers.heading":"Publish the common record|||protect only what needs it","trust-centre.documentation.accessLayers.body":"Public evidence should be discoverable without a contract. Protected evidence goes to regulators and authorised reviewers, and internal material never replaces the public record.","trust-centre.documentation.accessLayers.pills":"Public|||Scoped|||Protected","trust-centre.documentation.accessLayers.chrome":"Access model|||Evidence disclosure layers|||Four levels|||Least restriction with purpose","trust-centre.documentation.leastDisclosure":"Purpose boundary|||Access matches the evidence needed for the stated review","trust-centre.documentation.auditability":"Access record|||Protected disclosure remains attributable and versioned","trust-centre.documentation.confidentiality":"Necessary protection|||Personal data, security details, customer secrets, and intellectual property stay controlled","trust-centre.documentation.publicLayer":"Public|||Common facts, statuses, boundaries, policies, routes, methods and change history","trust-centre.documentation.downstreamLayer":"Downstream|||Interface, integration, evaluation context, limitations, configuration, and update information","trust-centre.documentation.regulatorLayer":"Regulator or authorised assessor|||Case-specific technical evidence under an appropriate protected route","trust-centre.documentation.internalLayer":"Internal|||Build, operations, security, decision, and investigation material with named ownership","trust-centre.documentation.environment.eyebrow":"Environmental evidence","trust-centre.documentation.environment.heading":"Publish the estimate|||publish its limits with it","trust-centre.documentation.environment.body":"Historical training was not metered per job, so Dweve publishes a reconstructed order-of-magnitude estimate rather than false precision. New specialists and releases will record wall time, component-hours, utilisation, workload, software version, captured state and energy where measurable.","trust-centre.documentation.environment.pills":"Hardware|||Method|||Uncertainty","trust-centre.documentation.environment.chrome":"Measurement schema|||Environmental record|||Six fields|||No borrowed benchmark","trust-centre.documentation.noEstimate":"No false precision|||The historical figure is explicitly an order-of-magnitude estimate, not a metered total or comparative environmental claim","trust-centre.documentation.methodRequired":"Method disclosed|||Elapsed time, duty range, node-equivalent range, mixed draw and allocated storage/network overhead are stated with the estimate","trust-centre.documentation.releaseGate":"Prospective measurement contract|||New specialists and releases record their matching Loom state, workload, software, component-hours and energy where measurable","trust-centre.documentation.hardware":"Dweve training estate|||ASUS and Supermicro servers; AMD EPYC CPUs; AMD Xilinx Alveo U55C FPGA; NVIDIA L40S GPU; Kingston ECC memory; Solidigm NVMe; Mellanox 100GbE. Quantities are not public. No fixed customer hardware is required.","trust-centre.documentation.energyValue":"Historical training energy|||Best estimate 18 MWh; plausible range 4 to 50 MWh; low confidence and not contemporaneously metered","trust-centre.documentation.measurement":"Reconstruction method|||21,900 elapsed hours × 20 to 50% active duty × one to three compute-node equivalents × 0.8 to 1.3 kW mixed CPU/FPGA/GPU draw, plus 15% allocated storage and network overhead","trust-centre.documentation.period":"Training period|||Approximately 2.5 years, or 21,900 elapsed hours, of intermittent machine work inside a longer human-in-the-loop research and verification programme from 2023 to the present","trust-centre.documentation.uncertainty":"Uncertainty|||Low confidence: historical per-job telemetry was not retained, work was intermittent, concurrency varied, and the estimate must not be compared with differently scoped training figures as if they were equivalent","trust-centre.documentation.scopeBoundary":"Lifecycle boundary|||Includes reconstructed compute-node draw and 15% allocated storage/network overhead; excludes building energy and embodied carbon. Inference is route-shaped and therefore has no single representative value.","trust-centre.case.topic.model-behaviour":"Model behaviour","trust-centre.case.topic.model-behaviour.short":"Unexpected, incorrect, or unsafe output","trust-centre.case.topic.model-behaviour.body":"A recorded route for unexpected, incorrect, harmful, or unsafe Loom behaviour.","trust-centre.case.topic.copyright":"Copyright and rights","trust-centre.case.topic.copyright.short":"Works, reservations, crawling, or removal","trust-centre.case.topic.copyright.body":"A structured rightsholder route for works, rights reservations, crawler activity, exclusion, and removal.","trust-centre.case.topic.security":"Security vulnerability","trust-centre.case.topic.security.short":"A weakness in a Dweve surface","trust-centre.case.topic.security.body":"A coordinated disclosure route for a vulnerability affecting Dweve software, services, or infrastructure.","trust-centre.case.topic.serious-incident":"Serious AI incident","trust-centre.case.topic.serious-incident.short":"Actual or imminent serious harm","trust-centre.case.topic.serious-incident.body":"An incident route for actual or imminent serious harm involving a Dweve system.","trust-centre.case.topic.privacy":"Privacy and data","trust-centre.case.topic.privacy.short":"Personal data, retention, access, or feedback","trust-centre.case.topic.privacy.body":"A recorded route for personal data, retention, access, deletion, feedback, or data-boundary concerns.","trust-centre.case.topic.downstream":"Downstream documentation","trust-centre.case.topic.downstream.short":"Integration evidence or controlled pack","trust-centre.case.topic.downstream.body":"A documentation request for an integrator, deployer, procurement team, authority, or assessor.","trust-centre.case.topic.other":"Other Trust matter","trust-centre.case.topic.other.short":"A matter that does not fit another route","trust-centre.case.topic.other.body":"Trust operations will triage the report and record the accountable owner.","trust-centre.case.form.topic":"Choose the accountable route","trust-centre.case.form.route":"Selected route","trust-centre.case.form.evidence":"Topic-specific evidence","trust-centre.case.form.consent":"I ask Dweve to process the information in this report to investigate, respond, preserve an audit record, and meet applicable legal obligations. I understand that the private status link is the access key to this case.","trust-centre.case.form.sending":"Creating case","trust-centre.case.form.submit":"Create recorded case","trust-centre.case.form.error":"Check the marked required fields and try again. Nothing has been recorded yet.","trust-centre.case.form.audit":"Submission creates a case ID, owner, receipt event, and private status route.","trust-centre.case.form.existing":"Open an existing case","trust-centre.case.field.contactName":"Contact name","trust-centre.case.field.email":"Email address","trust-centre.case.field.organisation":"Organisation","trust-centre.case.field.role":"Role or relationship","trust-centre.case.field.summary":"Short case summary","trust-centre.case.field.description":"Complete report","trust-centre.case.field.description.placeholder":"Describe what happened, what you observed, why it matters, and what evidence can establish it.","trust-centre.case.field.evidenceUrls":"Supporting links, one per line","trust-centre.case.field.evidenceUrls.placeholder":"https://example.eu/evidence","trust-centre.case.field.claimantCapacity":"Your capacity","trust-centre.case.field.claimantCapacity.placeholder":"Rightsholder, authorised representative, collective management organisation, or another relationship","trust-centre.case.field.workDescription":"Work or protected material","trust-centre.case.field.workDescription.placeholder":"Identify the work, collection, right, or protected material precisely.","trust-centre.case.field.rightsBasis":"Rights basis and authority","trust-centre.case.field.rightsBasis.placeholder":"Explain the rights you hold or the authority under which you act.","trust-centre.case.field.locationUrls":"Locations concerned, one URL per line","trust-centre.case.field.locationUrls.placeholder":"https://example.eu/work-or-reservation","trust-centre.case.field.requestedAction":"Requested action","trust-centre.case.field.requestedAction.placeholder":"State the investigation, exclusion, correction, removal, preservation, or response you request.","trust-centre.case.field.integrationPurpose":"Integration purpose","trust-centre.case.field.integrationPurpose.placeholder":"Describe the system, users, decisions, and deployment context in which Loom would be integrated.","trust-centre.case.field.deploymentRoute":"Planned access or deployment route","trust-centre.case.field.deploymentRoute.placeholder":"Fabric, Aura, API, licensed local deployment, Mesh, or a combination","trust-centre.case.field.documentationRequested":"Documentation requested","trust-centre.case.field.documentationRequested.placeholder":"List the model, integration, evaluation, limitation, data, security, or change records needed.","trust-centre.case.field.confidentialityRoute":"Confidentiality route","trust-centre.case.field.confidentialityRoute.placeholder":"Public documents only, NDA available, existing agreement, or another controlled route","trust-centre.case.field.interactionRoute":"Interaction route","trust-centre.case.field.interactionRoute.placeholder":"Fabric, Aura, API, local deployment, Mesh, or another route","trust-centre.case.field.occurredAt":"When it occurred","trust-centre.case.field.occurredAt.placeholder":"Date and time if known","trust-centre.case.field.outputReference":"Output or trace reference","trust-centre.case.field.outputReference.placeholder":"Provide a trace ID, receipt, screenshot description, output excerpt, or another reproducible reference.","trust-centre.case.field.expectedBehaviour":"Expected behaviour","trust-centre.case.field.expectedBehaviour.placeholder":"Explain the expected safe or correct behaviour and how the observed result differed.","trust-centre.case.field.vulnerabilityType":"Vulnerability type","trust-centre.case.field.vulnerabilityType.placeholder":"For example access control, injection, data exposure, supply chain, or cryptographic integrity","trust-centre.case.field.affectedSurface":"Affected surface and version","trust-centre.case.field.affectedSurface.placeholder":"Identify the product, endpoint, package, version, configuration, or deployment boundary.","trust-centre.case.field.reproductionSteps":"Reproduction steps","trust-centre.case.field.reproductionSteps.placeholder":"Give the smallest safe sequence that reproduces the weakness. Do not include live secrets.","trust-centre.case.field.impact":"Observed or plausible impact","trust-centre.case.field.impact.placeholder":"Explain affected confidentiality, integrity, availability, safety, or users.","trust-centre.case.field.affectedPeopleOrSystems":"People or systems affected","trust-centre.case.field.affectedPeopleOrSystems.placeholder":"Describe who or what was affected and the known scale.","trust-centre.case.field.immediateRisk":"Immediate or continuing risk","trust-centre.case.field.immediateRisk.placeholder":"Explain whether harm is continuing, imminent, contained, or unknown.","trust-centre.case.field.authoritiesNotified":"Authorities or responders already notified","trust-centre.case.field.authoritiesNotified.placeholder":"Optional: name any authority, emergency service, operator, or responder already informed.","trust-centre.case.field.privacyContext":"Data and privacy context","trust-centre.case.field.privacyContext.placeholder":"Describe the data, person, processing route, request, retention, or feedback issue involved.","trust-centre.case.receipt.eyebrow":"Recorded immediately","trust-centre.case.receipt.title":"Your case now has an identity","trust-centre.case.receipt.body":"The receipt, routing event, responsible team, and status history are stored together. A copy and the private status link are sent to the supplied email address.","trust-centre.case.receipt.id":"Case ID","trust-centre.case.receipt.owner":"Responsible team","trust-centre.case.receipt.ownerPending":"Trust operations","trust-centre.case.receipt.status":"Current status","trust-centre.case.receipt.target":"Response target","trust-centre.case.receipt.open":"Open private case status","trust-centre.case.receipt.private":"The status URL contains a private access token. Treat it like a confidential case link.","trust-centre.case.status.received":"Received","trust-centre.case.status.triage":"In triage","trust-centre.case.status.in-review":"Under review","trust-centre.case.status.waiting-on-reporter":"Waiting on reporter","trust-centre.case.status.resolved":"Resolved","trust-centre.case.status.closed":"Closed","trust-centre.case.glyph.live":"Recorded route","trust-centre.case.glyph.receive":"Receive","trust-centre.case.glyph.receive.body":"Validate and preserve the report","trust-centre.case.glyph.identify":"Identify","trust-centre.case.glyph.identify.body":"Issue the permanent case ID","trust-centre.case.glyph.route":"Route","trust-centre.case.glyph.route.body":"Assign the accountable team","trust-centre.case.glyph.track":"Track","trust-centre.case.glyph.track.body":"Publish status events privately","trust-centre.case.routeNav.label":"Trust reporting routes","trust-centre.case.routeNav.copyright":"Rightsholders","trust-centre.case.routeNav.downstream":"Documentation","trust-centre.case.routeNav.model":"Model behaviour","trust-centre.case.routeNav.security":"Security","trust-centre.case.routeNav.incident":"Serious incident","trust-centre.case.assurance.title":"What the route creates","trust-centre.case.assurance.caseId":"Stable case ID","trust-centre.case.assurance.caseId.body":"A human-readable identifier for every follow-up","trust-centre.case.assurance.audit":"Append-only events","trust-centre.case.assurance.audit.body":"Receipt, routing, status, and public responses remain ordered","trust-centre.case.assurance.private":"Private access key","trust-centre.case.assurance.private.body":"The reporter controls access to the case-status view","trust-centre.case.assurance.response":"Accountable response","trust-centre.case.assurance.response.body":"The route records an owner and any applicable response target","trust-centre.case.page.start":"Start the report","trust-centre.case.page.status":"Check a case","trust-centre.case.page.form.pill.receipt":"Immediate receipt","trust-centre.case.page.form.pill.owner":"Named owner","trust-centre.case.page.form.pill.status":"Private status history","trust-centre.case.page.form.pill.fourteen":"Fourteen-day response","trust-centre.case.page.copyright.eyebrow":"Rightsholder channel","trust-centre.case.page.copyright.lead":"Name the work and the right","trust-centre.case.page.copyright.accent":"then follow the case","trust-centre.case.page.copyright.lede":"A structured route for rights reservations, crawler activity, source inclusion, exclusion, correction, and removal. The receipt is a real case, not an email disappearing into a shared inbox.","trust-centre.case.page.copyright.glyph":"Rights case workflow","trust-centre.case.page.copyright.crumb":"Rightsholder report","trust-centre.case.page.copyright.form.eyebrow":"Structured rights evidence","trust-centre.case.page.copyright.form.lead":"Identify what is protected","trust-centre.case.page.copyright.form.accent":"and what should happen","trust-centre.case.page.copyright.form.body":"Provide the work, your authority, relevant locations, the requested action, and supporting material. Dweve records the receipt, route, case owner, and response history.","trust-centre.case.page.copyright.form.footer":"RIGHTSHOLDER CASE / RECEIPT + ROUTING + STATUS","trust-centre.case.page.downstream.eyebrow":"Downstream documentation","trust-centre.case.page.downstream.lead":"Describe the integration","trust-centre.case.page.downstream.accent":"receive the right evidence pack","trust-centre.case.page.downstream.lede":"Public records stay open. Controlled integration material is scoped to the intended system, deployment route, confidentiality boundary, and Loom version. Dweve responds within fourteen calendar days.","trust-centre.case.page.downstream.glyph":"Documentation request","trust-centre.case.page.downstream.crumb":"Request documentation","trust-centre.case.page.downstream.form.eyebrow":"Scoped integration request","trust-centre.case.page.downstream.form.lead":"Tell us what you are building","trust-centre.case.page.downstream.form.accent":"and what must be evidenced","trust-centre.case.page.downstream.form.body":"The request records purpose, deployment route, requested documents, contact role, confidentiality route, and the fourteen-day response target.","trust-centre.case.page.downstream.form.footer":"DOWNSTREAM PACK / FOURTEEN CALENDAR DAYS","trust-centre.case.page.model.eyebrow":"Model behaviour report","trust-centre.case.page.model.lead":"Preserve the output context","trust-centre.case.page.model.accent":"make the behaviour reviewable","trust-centre.case.page.model.lede":"Report an unsafe, incorrect, unexpected, or harmful Loom result with the access route, time, trace or output reference, expected behaviour, and supporting evidence.","trust-centre.case.page.model.glyph":"Behaviour review route","trust-centre.case.page.model.crumb":"Report model behaviour","trust-centre.case.page.model.form.eyebrow":"Reproducible report","trust-centre.case.page.model.form.lead":"Capture what Loom did","trust-centre.case.page.model.form.accent":"and what should have happened","trust-centre.case.page.model.form.body":"A trace reference helps, but is not required. Give enough context to reproduce or bound the behaviour without exposing unrelated personal or confidential data.","trust-centre.case.page.model.form.footer":"MODEL BEHAVIOUR / VERSION + ROUTE + EVIDENCE","trust-centre.case.page.security.eyebrow":"Coordinated disclosure","trust-centre.case.page.security.lead":"Report the weakness safely","trust-centre.case.page.security.accent":"track the response privately","trust-centre.case.page.security.lede":"Identify the affected surface and version, provide the smallest safe reproduction, and explain the impact. Do not submit live credentials or exploit unrelated systems.","trust-centre.case.page.security.glyph":"Security disclosure route","trust-centre.case.page.security.crumb":"Report a vulnerability","trust-centre.case.page.security.form.eyebrow":"Security evidence","trust-centre.case.page.security.form.lead":"Bound the affected surface","trust-centre.case.page.security.form.accent":"show the reproducible path","trust-centre.case.page.security.form.body":"The case is routed to Security and retains the receipt, version context, public status events, and response. Sensitive exploit detail remains in the protected case record.","trust-centre.case.page.security.form.footer":"COORDINATED DISCLOSURE / PROTECTED CASE","trust-centre.case.page.incident.eyebrow":"Serious incident route","trust-centre.case.page.incident.lead":"Record actual or imminent harm","trust-centre.case.page.incident.accent":"route it without delay","trust-centre.case.page.incident.lede":"Use this route when a Dweve system may be connected to death, serious injury, severe service disruption, critical-infrastructure impact, fundamental-rights harm, or another serious outcome.","trust-centre.case.page.incident.glyph":"Incident response route","trust-centre.case.page.incident.crumb":"Report a serious incident","trust-centre.case.page.incident.form.eyebrow":"Incident evidence","trust-centre.case.page.incident.form.lead":"State who or what is affected","trust-centre.case.page.incident.form.accent":"and whether the risk continues","trust-centre.case.page.incident.form.body":"Give the occurrence time, affected people or systems, immediate risk, known containment, other responders, and supporting evidence. Contact emergency services first where immediate physical safety requires it.","trust-centre.case.page.incident.form.footer":"SERIOUS INCIDENT / PRESERVE + ROUTE + TRACK","trust-centre.case.statusPage.eyebrow":"Private case status","trust-centre.case.statusPage.lead":"One case identity","trust-centre.case.statusPage.accent":"one ordered response history","trust-centre.case.statusPage.lede":"Open the private case view with the case ID and access token from the acknowledgement. The view contains status events and public responses, never internal investigation notes.","trust-centre.case.statusPage.crumb":"Case status","trust-centre.case.statusPage.glyph":"Private status access","trust-centre.case.statusPage.private":"Token-protected case view","trust-centre.case.statusPage.jump":"Open case status","trust-centre.case.statusPage.report":"Create a new report","trust-centre.case.statusPage.section.eyebrow":"Status lookup","trust-centre.case.statusPage.section.lead":"Use the receipt identity","trust-centre.case.statusPage.section.accent":"not an inbox search","trust-centre.case.statusPage.section.body":"The case ID identifies the record. The private access token proves possession of the acknowledgement link. Dweve stores only the token hash.","trust-centre.case.statusPage.pill.1":"Case ID","trust-centre.case.statusPage.pill.2":"Hashed access token","trust-centre.case.statusPage.pill.3":"Public event history","trust-centre.case.statusPage.section.footer":"PRIVATE STATUS / NO INTERNAL NOTES","trust-centre.case.statusPage.id":"Case ID","trust-centre.case.statusPage.token":"Private access token","trust-centre.case.statusPage.loading":"Opening case","trust-centre.case.statusPage.open":"Open case","trust-centre.case.statusPage.error":"The case ID and private token do not match an accessible case.","trust-centre.case.statusPage.help":"Use the values from the acknowledgement email. The token is never displayed in the public case record.","trust-centre.case.statusPage.record":"Case record","trust-centre.case.statusPage.topic":"Route","trust-centre.case.statusPage.model":"Applies to","trust-centre.case.statusPage.owner":"Responsible team","trust-centre.case.statusPage.updated":"Last public update","trust-centre.case.statusPage.response":"Latest public response","trust-centre.case.statusPage.new":"Create another report","trust-centre.case.event.receipt-issued":"Receipt issued","trust-centre.case.event.status-updated":"Status updated","trust-centre.record.label":"Versioned public record","trust-centre.record.id":"Stable record ID","trust-centre.record.version":"Content version","trust-centre.record.status":"Status","trust-centre.record.status.current":"Current","trust-centre.record.status.prerelease":"Pre-release","trust-centre.record.status.pending":"Future event","trust-centre.record.applies":"Applies to","trust-centre.record.applies.loom":"Dweve Loom 1.0","trust-centre.record.applies.trust":"Dweve Trust system","trust-centre.record.effective":"Effective from","trust-centre.record.verified":"Last reviewed against source record","trust-centre.record.owner":"Owner and contact","trust-centre.record.links":"Record downloads and history","trust-centre.record.machine":"Machine record","trust-centre.record.previous":"Previous version","trust-centre.record.integrity":"SHA-256 manifest","trust-centre.statutory.downloads":"Canonical record and exports","trust-centre.statutory.model.eyebrow":"Commission model documentation structure","trust-centre.statutory.model.title":"One model record with every required field visible","trust-centre.statutory.model.body":"This schedule follows the European Commission Model Documentation Form. Confirmed facts, bounded estimates and measurement limitations are stated directly, with controlled evidence routed to the appropriate access layer.","trust-centre.statutory.model.group.general":"General information","trust-centre.statutory.model.group.general.body":"Provider, model identity, authenticity, release, Union placement and dependencies","trust-centre.statutory.model.group.properties":"Model properties","trust-centre.statutory.model.group.properties.body":"Architecture, design, modalities, limits and total size","trust-centre.statutory.model.group.distribution":"Distribution and licences","trust-centre.statutory.model.group.distribution.body":"Current and planned channels, access level, licence and accompanying assets","trust-centre.statutory.model.group.use":"Use and integration","trust-centre.statutory.model.group.use.body":"Intended and prohibited uses, integrating systems and deployment requirements","trust-centre.statutory.model.group.training":"Training process","trust-centre.statutory.model.group.training.body":"Training design and the rationale for the consequential choices","trust-centre.statutory.model.group.data":"Training, testing and validation data","trust-centre.statutory.model.group.data.body":"Provenance, scale, scope, curation, unsuitability and identifiable bias","trust-centre.statutory.model.group.resources":"Compute and energy","trust-centre.statutory.model.group.resources.body":"Training duration, computation, energy and inference measurement","trust-centre.statutory.model.field.provider":"Legal provider and contact","trust-centre.statutory.model.field.provider.value":"Dweve B.V. · info@dweve.com · Arnhem, the Netherlands","trust-centre.statutory.model.field.model":"Versioned model name","trust-centre.statutory.model.field.model.value":"Dweve Loom 1.0","trust-centre.statutory.model.field.authenticity":"Model authenticity","trust-centre.statutory.model.field.authenticity.value":"Dweve Loom 1.0 is adaptive during use. Time-bound captured states and sealed deployment packages receive their own manifests and package-specific integrity evidence.","trust-centre.statutory.model.field.release":"First release date","trust-centre.statutory.model.field.release.value":"Internal invite-only pre-release is available; no external release has occurred as of 1 August 2026","trust-centre.statutory.model.field.unionRelease":"Union market release date","trust-centre.statutory.model.field.unionRelease.value":"Invite-only external access is planned for 1 September 2026; the actual placement date will be recorded when it occurs","trust-centre.statutory.model.field.dependencies":"Model dependencies","trust-centre.statutory.model.field.dependencies.value":"No external GPAI model dependency. Loom uses Dweve Core for execution, Spindle for governed constraints and provenance, BitWeave for retrieval, Numerus for exact numerics, AION for proof and replay, and Lattice for policy. Winnow serves source acquisition; Mesh and Selvedge are optional deployment infrastructure, not model dependencies.","trust-centre.statutory.model.field.retention":"Documentation retention","trust-centre.statutory.model.field.retention.value":"This version and every superseded model-documentation version are retained for at least 10 years after Union market placement.","trust-centre.statutory.model.field.architecture":"Model architecture","trust-centre.statutory.model.field.architecture.value":"528 top-level domain specialists with deeper sub-specialists, solvers, perception, memory, constraints and verification components. Dweve records Loom as a general-purpose AI model under Article 3(63); the Commission's 10²³-FLOP criterion is indicative, not a safe harbour for another architecture or representation.","trust-centre.statutory.model.field.design":"Design specifications","trust-centre.statutory.model.field.design.value":"Adaptive task-shaped routing; approximately 4 to 8 top-level domain specialists on average, without a hard minimum or maximum","trust-centre.statutory.model.field.inputs":"Input modalities","trust-centre.statutory.model.field.inputs.value":"Interprets all represented modalities, including language, code, structured data, images, audio, video and sensor data","trust-centre.statutory.model.field.outputs":"Output modalities","trust-centre.statutory.model.field.outputs.value":"Documented release outputs are text and code, plus the images Dweve ships; Loom does not generate video","trust-centre.statutory.model.field.sizes":"Maximum input and output sizes","trust-centre.statutory.model.field.sizes.value":"1,000,000 input units and 128,000 output units per request","trust-centre.statutory.model.field.parameters":"Total model size","trust-centre.statutory.model.field.parameters.value":"Approximately 1.5 trillion explicit domain constraints across 528 top-level domain specialists, about 3 billion per specialist. This is a constraint-state scale, not a dense neural parameter count; learned components are separate, sparse and task-selected.","trust-centre.statutory.model.field.currentChannel":"Current distribution","trust-centre.statutory.model.field.currentChannel.value":"Controlled internal pre-release testing","trust-centre.statutory.model.field.plannedChannels":"Planned invite-only channels","trust-centre.statutory.model.field.plannedChannels.value":"Fabric web, Aura, API, licensed local deployment and Mesh, individually or in combination","trust-centre.statutory.model.field.license":"Model licence","trust-centre.statutory.model.field.license.value":"Proprietary commercial access under the Dweve master terms, applicable product terms and signed order; the public licensing boundary is published at /trust/terms#licensing-terms","trust-centre.statutory.model.field.assets":"Additional assets","trust-centre.statutory.model.field.assets.value":"System card, Commission training summary, 772-entry source catalogue, admission policy, acceptable-use record, evaluation and monitoring records, controlled downstream pack, signed deployment packages and captured-state manifests","trust-centre.statutory.model.field.aup":"Acceptable Use Policy","trust-centre.statutory.model.field.aup.value":"Published at /trust/acceptable-use and versioned through the Trust change ledger","trust-centre.statutory.model.field.intended":"Intended and restricted uses","trust-centre.statutory.model.field.intended.value":"Governed assistance, analysis, interpretation, generation, orchestration and decision support within the published boundaries","trust-centre.statutory.model.field.systems":"Types of integrating systems","trust-centre.statutory.model.field.systems.value":"Fabric web, Aura, API integrations, licensed local deployments and Mesh deployments","trust-centre.statutory.model.field.integration":"Technical integration means","trust-centre.statutory.model.field.integration.value":"Public foundation plus a controlled, version-scoped documentation request with a fourteen-day response target","trust-centre.statutory.model.field.hardware":"Required hardware","trust-centre.statutory.model.field.hardware.value":"No fixed customer hardware is required. Dweve's provider estate uses ASUS ESC4000A-E12 compute servers, AMD EPYC 9454P and 9354P processors, Kingston DDR5 ECC memory, AMD Xilinx Alveo U55C FPGAs, NVIDIA L40S GPUs, Mellanox ConnectX-6 Dx and SN3700 100GbE networking, Solidigm D7-P5520 NVMe and Supermicro AS-2115HS-TNR storage. Non-visual training is native binary/integer; GPUs are still used to train visual modalities while Dweve researches a binary visual-training path. Unit counts are not public.","trust-centre.statutory.model.field.software":"Required software","trust-centre.statutory.model.field.software.value":"Only Dweve software is required: the signed Dweve runtime and the Dweve components selected for the deployment. No third-party model or external model service is required to run Loom.","trust-centre.statutory.model.field.trainingDesign":"Training design specifications","trust-centre.statutory.model.field.trainingDesign.value":"Dweve extracts and deduces explicit constraints and relationships from available datasets and other sources, including Hugging Face and permitted web sources. Loom's own calculations can propose additional candidates; Spindle researches, cross-checks and validates them before storage, so governed constraint sets expand. Candidate constraints are checked across multiple sources and repeated occurrences with contradiction and domain checks; permitted-crawling controls, source and rights evidence, and human-in-the-loop review for part of the process are retained. User data is not raw training material or silently mined into constraints. When Loom explicitly asks a user a question because governed state cannot answer, it tells the user that the answer will train the governed constraint system; only a derived candidate may enter the same Spindle route. Rights, access, provenance, privacy and purpose gates determine which records may be promoted; they govern admission, not the origin of candidate constraints. Promoted constraints retain evidence and lineage in Spindle. Small bipolar transformer components render completed result graphs into coherent language; they do not determine the result.","trust-centre.statutory.model.field.rationale":"Training decision rationale","trust-centre.statutory.model.field.rationale.value":"Separating source-backed knowledge, explicit constraints, solvers, learned components, verification and final expression makes each responsibility inspectable, correctable and versionable instead of hiding it in one undifferentiated parameter mass.","trust-centre.statutory.model.field.dataModalities":"Data types and modalities","trust-centre.statutory.model.field.dataModalities.value":"The source catalogue covers natural language, code, structured and tabular data, formal mathematics, images, audio, video, sensor and scientific measurements, geospatial records, simulations and synthetic supervision. Each admitted record is assigned to constraint extraction, language rendering, evaluation or exclusion.","trust-centre.statutory.model.field.provenance":"Data provenance route","trust-centre.statutory.model.field.provenance.value":"Winnow records source acquisition and robots, rate and target-safety controls. Spindle binds source, item, licence, attribution, purpose, evidence and promotion state. Loom consumes the governed constraint and component state selected for its captured state.","trust-centre.statutory.model.field.selection":"Acquisition and selection","trust-centre.statutory.model.field.selection.value":"ALLOW requires a catalogued source, approved effective licence, commercial and derivative rights, source and item provenance, captured attribution, resolved third-party payload, and passed privacy and confidentiality gates. A critical UNKNOWN is quarantined, never allowed.","trust-centre.statutory.model.field.dataPoints":"Number of data points","trust-centre.statutory.model.field.dataPoints.value":"772 disclosed source-family entries across 19 groups feed record-level admission. The top-level domain specialist catalogue holds approximately 1.5 trillion explicit domain constraints, about 3 billion per specialist. Raw source-item and modality counts remain in the internal Spindle ledger rather than being misrepresented as one token total.","trust-centre.statutory.model.field.scope":"Scope and main characteristics","trust-centre.statutory.model.field.scope.value":"Multilingual public and licensed material spanning law, government, scholarship, culture, code, standards, mathematics, physics, chemistry, biology, medicine, climate, geography, language and speech, knowledge graphs, economics, agriculture, media, education and engineering.","trust-centre.statutory.model.field.curation":"Data curation methods","trust-centre.statutory.model.field.curation.value":"Discovery across available datasets and other sources, including Hugging Face and permitted web sources; source-version capture; extraction and deduction of constraints and relationships; candidate generation from Loom's own calculations; cross-source and repeated-occurrence corroboration; contradiction and domain checks; licence, provenance and attribution resolution; reciprocal-lineage isolation; deduplication; source validation; conflict handling; evidence attachment; human-in-the-loop review for part of the process; and versioned Spindle promotion. Candidates derived from an explicitly asked-for user answer follow the same disclosed research and validation route, while raw user data is not training material. Rights and access gates govern promotion rather than the origin of candidate constraints.","trust-centre.statutory.model.field.unsuitable":"Unsuitable-source detection","trust-centre.statutory.model.field.unsuitable.value":"Missing provenance, unresolved rights, incompatible terms, third-party payloads, privacy or confidentiality failures, access-control bypass, illegal material and validation failure lead to quarantine or exclusion. Filterable families must pass at item level; rights-instrument sources require the archived instrument and version.","trust-centre.statutory.model.field.bias":"Identifiable-bias measures","trust-centre.statutory.model.field.bias.value":"Spindle preserves source, language, geography, domain, licence class and evidence lineage so coverage imbalance and conflicting rules remain inspectable. Specialist-specific evaluation, contradiction detection, belief revision and source challenges test the promoted state; no claim of universal neutrality is made.","trust-centre.statutory.model.field.trainingTime":"Training time","trust-centre.statutory.model.field.trainingTime.value":"Approximately 2.5 years of intermittent machine training, constraint extraction and validation, inside a longer multi-year human-in-the-loop research and verification programme. This is about 21,900 elapsed hours, not 21,900 continuously loaded compute hours.","trust-centre.statutory.model.field.trainingCompute":"Training computation","trust-centre.statutory.model.field.trainingCompute.value":"Estimated 4,380 to 32,850 active compute-node-equivalent hours; central estimate 15,330. The range reflects intermittent 20 to 50% duty and one to three simultaneous compute-node equivalents. Historical work was performed on and off across the disclosed Dweve estate; fleet quantities remain confidential.","trust-centre.statutory.model.field.computeMethod":"Compute measurement method","trust-centre.statutory.model.field.computeMethod.value":"Reconstructed estimate: 21,900 elapsed hours multiplied by 20 to 50% active duty and one to three simultaneous compute-node equivalents. Historical job-level telemetry was not retained. New domain specialists and model releases will record wall time, component-hours, utilisation, workload, software version and captured-state identity from the start.","trust-centre.statutory.model.field.trainingEnergy":"Training energy","trust-centre.statutory.model.field.trainingEnergy.value":"Best estimate approximately 18 MWh; plausible historical range 4 to 50 MWh. This is an order-of-magnitude estimate, not a metered total. It covers mixed CPU, FPGA and GPU training plus allocated storage and network overhead.","trust-centre.statutory.model.field.energyMethod":"Energy measurement method","trust-centre.statutory.model.field.energyMethod.value":"Active compute-node-equivalent hours multiplied by an estimated 0.8 to 1.3 kW mixed training draw, plus 15% allocated storage and network overhead. CPU, memory, FPGA, GPU, NVMe, NIC and chassis load are inside the boundary; building energy and embodied carbon are outside it. Uncertainty is high because historical work was not metered per job.","trust-centre.statutory.model.field.inferenceCompute":"Benchmarked inference computation","trust-centre.statutory.model.field.inferenceCompute.value":"No single meaningful value: Loom activates a task-shaped graph and typically selects about 4 to 8 top-level domain specialists, without a hard minimum or maximum. Request-level computation depends on modality, evidence retrieval, solvers, verification and available resources.","trust-centre.statutory.model.field.inferenceMethod":"Inference measurement method","trust-centre.statutory.model.field.inferenceMethod.value":"From external beta onward, route receipts will bind active specialists and components, operation counts, wall time, device class, energy where measurable, software versions and captured state. Reports will publish workload-equivalent distributions rather than one misleading average.","trust-centre.statutory.training.eyebrow":"Commission public training summary structure","trust-centre.statutory.training.title":"The Commission structure, source catalogue, and open evidence fields","trust-centre.statutory.training.body":"Every field in the Commission template is present. The public record states the 772 source families, admission classes, collection period, modalities, rights controls and processing route directly. Exact admitted-item decisions and provenance stay in the controlled Spindle ledger so the public summary remains useful without exposing an operational evidence store.","trust-centre.statutory.training.group.provider":"Provider identification","trust-centre.statutory.training.group.provider.body":"EU-established provider, public contact and representative status","trust-centre.statutory.training.group.model":"Model identification","trust-centre.statutory.training.group.model.body":"Versioned model, dependencies and date of Union placement","trust-centre.statutory.training.group.characteristics":"Modalities, size and characteristics","trust-centre.statutory.training.group.characteristics.body":"Training modalities, volume, content types, dates, languages and relevant characteristics","trust-centre.statutory.training.group.publicDatasets":"Publicly available datasets","trust-centre.statutory.training.group.publicDatasets.body":"Use, large datasets above the template threshold and the remaining public collections","trust-centre.statutory.training.group.privateDatasets":"Private third-party datasets","trust-centre.statutory.training.group.privateDatasets.body":"Commercially licensed collections and other non-public third-party sources","trust-centre.statutory.training.group.crawled":"Crawled and scraped online sources","trust-centre.statutory.training.group.crawled.body":"Crawler identity, purpose, behaviour, collection period, content and most relevant domains","trust-centre.statutory.training.group.userData":"User data","trust-centre.statutory.training.group.userData.body":"Ordinary interactions, other service data and separately submitted feedback","trust-centre.statutory.training.group.synthetic":"Synthetic data","trust-centre.statutory.training.group.synthetic.body":"Whether synthetic data was used and which models created it","trust-centre.statutory.training.group.otherSources":"Other data sources","trust-centre.statutory.training.group.otherSources.body":"Offline, self-digitised, commissioned human or other uncategorised sources","trust-centre.statutory.training.group.rights":"Reservation of TDM rights","trust-centre.statutory.training.group.rights.body":"Code commitment, honoured protocols and collection-stage rights controls","trust-centre.statutory.training.group.illegalContent":"Removal of illegal content","trust-centre.statutory.training.group.illegalContent.body":"Measures used to avoid or remove unlawful material under Union law","trust-centre.statutory.training.group.otherProcessing":"Other processing information","trust-centre.statutory.training.group.otherProcessing.body":"Voluntary processing detail and the governed route through Winnow, Spindle and Loom","trust-centre.statutory.training.field.providerIdentity":"Provider and contact details","trust-centre.statutory.training.field.providerIdentity.value":"Dweve B.V. · info@dweve.com · Arnhem, the Netherlands","trust-centre.statutory.training.field.representative":"Authorised representative","trust-centre.statutory.training.field.representative.value":"Not applicable because Dweve B.V. is established in the European Union","trust-centre.statutory.training.field.modelIdentity":"Versioned model name","trust-centre.statutory.training.field.modelIdentity.value":"Dweve Loom 1.0","trust-centre.statutory.training.field.dependencies":"Model dependencies","trust-centre.statutory.training.field.dependencies.value":"No external general-purpose model dependency. Loom uses Dweve Core, Spindle, BitWeave, Numerus, AION and Lattice as internal model/runtime components; Winnow, Reed, Ledger, FMI, Twin and Selvedge support governed acquisition, parsing, evidence, simulation, state validation and controlled execution.","trust-centre.statutory.training.field.marketDate":"Date placed on the Union market","trust-centre.statutory.training.field.marketDate.value":"Not yet occurred. Invite-only Union access is planned for 1 September 2026; the actual placement date will be recorded when it happens.","trust-centre.statutory.training.field.trainingModalities":"Training-data modalities","trust-centre.statutory.training.field.trainingModalities.value":"Natural language, code, structured and tabular records, formal mathematics and logic, images and other visual material, audio, sensor and scientific measurements, geospatial records, simulations and typed multimodal material. Video can be interpreted; Loom does not generate video.","trust-centre.statutory.training.field.trainingSizes":"Training-data size by modality","trust-centre.statutory.training.field.trainingSizes.value":"The governed knowledge state contains approximately 1.5 trillion explicit domain constraints across 528 top-level domain specialists, about 3 billion per specialist. The catalogue contains 772 source-family entries in 19 groups. Dweve did not train a single weighted token corpus, so historical byte or token totals by modality would not describe the system faithfully; future specialist releases will meter source records and modality volumes prospectively.","trust-centre.statutory.training.field.contentTypes":"Types of content","trust-centre.statutory.training.field.contentTypes.value":"Public-sector and legal records, scholarship, cultural works, software and specifications, formal proofs, scientific and medical data, climate and geographic records, language and speech resources, knowledge graphs, economic and social data, education, engineering, images, audio and evaluation material, as enumerated in the 19-group catalogue.","trust-centre.statutory.training.field.latestDate":"Latest acquisition date","trust-centre.statutory.training.field.latestDate.value":"Source collection and constraint mining began during Dweve R&D in 2023 and continue under governance. This summary is current to 1 August 2026. Spindle records the acquisition time and source version for each admitted record.","trust-centre.statutory.training.field.languages":"Linguistic characteristics","trust-centre.statutory.training.field.languages.value":"Multilingual by design, with Dutch, English and other European languages supported by official EU and national corpora, plus broad global-language public and licensed sources. Language, script, region and source provenance remain attached where the source provides them.","trust-centre.statutory.training.field.otherCharacteristics":"Other relevant characteristics","trust-centre.statutory.training.field.otherCharacteristics.value":"The primary training output is source-validated constraints and relationships rather than one opaque parameter mass. Candidate facts are corroborated repeatedly through source comparison, contradiction checks, domain validation and human-in-the-loop review before promotion. Small bipolar language components learn to express completed result graphs; they do not decide the result.","trust-centre.statutory.training.field.publicUse":"Public dataset use and modalities","trust-centre.statutory.training.field.publicUse.value":"Yes. Public-domain, CC0, permissive, attribution and approved reciprocal sources are used across every disclosed modality only after the applicable A0, A1, A2 or record-level F gate passes.","trust-centre.statutory.training.field.largeDatasets":"Large public datasets","trust-centre.statutory.training.field.largeDatasets.value":"Large aggregates are named in the 772-entry catalogue and subject to source, rights, access and corroboration gates. Dweve has no single token-mixture denominator from which a defensible three-percent share can be calculated; source-family counts are not presented as training weight.","trust-centre.statutory.training.field.otherPublic":"Other public datasets","trust-centre.statutory.training.field.otherPublic.value":"The complete public source-family list is published in the training-source inventory, including the licence class, applicable gate, usable slice and principal value for each entry.","trust-centre.statutory.training.field.licensed":"Commercially licensed source families","trust-centre.statutory.training.field.licensed.value":"Yes, as one source class, not as the origin of every constraint. The catalogue identifies 34 R-class source families that require a named registration, affiliate, click-through, API, territorial or commercial rights instrument. Other admitted material comes through public, attribution, reciprocal-open or filterable-source routes. The exact instrument and version are archived before use.","trust-centre.statutory.training.field.otherPrivate":"Other private third-party datasets","trust-centre.statutory.training.field.otherPrivate.value":"User prompts, conversations, files, private workspaces and customer documents are not raw training material and are not silently mined into constraints. If Loom explicitly asks a knowledge-gap question, it tells the user that the answer will train the governed constraint system; only a derived candidate may enter the disclosed Spindle route. Any non-public third-party source must be catalogued under an R-class rights instrument and pass the same provenance, privacy, confidentiality and purpose gates; there is no informal private-data route.","trust-centre.statutory.training.field.crawler":"Crawler name and identifier","trust-centre.statutory.training.field.crawler.value":"Winnow · open-source Apache 2.0 implementation at github.com/dweve-ai/winnow · public identifier Dweve-Winnow/1.0 (+https://dweve.com/trust/copyright)","trust-centre.statutory.training.field.purpose":"Crawler purpose","trust-centre.statutory.training.field.purpose.value":"Acquire candidate material for governed training workflows and recorded Loom tool retrieval","trust-centre.statutory.training.field.behaviour":"Crawler behaviour","trust-centre.statutory.training.field.behaviour.value":"Deny by default, evaluate robots.txt and source policy, apply per-domain rates, block internal targets and retain audit records","trust-centre.statutory.training.field.period":"Collection period","trust-centre.statutory.training.field.period.value":"2023 to the present, with continuous governed source discovery. The current summary closes on 1 August 2026; each request and admitted record carries its own timestamp.","trust-centre.statutory.training.field.crawledContent":"Crawled content and modalities","trust-centre.statutory.training.field.crawledContent.value":"Web pages, search results, APIs, feeds, repositories, legislation, public datasets, academic and government sources, CKAN/Socrata portals, SPARQL and CSV endpoints, streams and permitted files. Modalities include text, code, structured data, images, audio and other typed records. Winnow acquires and Reed parses; BitWeave retrieves; Spindle and Lattice govern; Ledger and AION preserve and verify evidence; FMI/Twin validate behaviour and state; Selvedge controls executable material.","trust-centre.statutory.training.field.domains":"Most relevant domains","trust-centre.statutory.training.field.domains.value":"The 772-entry catalogue spans 558 distinct domains. Its 28 most represented domains (top 5% by catalogue entries, not crawl traffic) are github.com, huggingface.co, ebi.ac.uk, ncbi.nlm.nih.gov, dumps.wikimedia.org, openslr.org, allenai.org, fda.gov, nlm.nih.gov, archive.org, data.europa.eu, fao.org, obofoundry.org, tensorflow.org, usgs.gov, cds.climate.copernicus.eu, commons.wikimedia.org, creativecommons.org, crystallography.net, dnb.nl, earthdata.nasa.gov, ec.europa.eu, gutenberg.org, gwosc.org, hplt-project.org, iana.org, joint-research-centre.ec.europa.eu and loc.gov.","trust-centre.statutory.training.field.ordinaryInteractions":"Ordinary model interactions","trust-centre.statutory.training.field.ordinaryInteractions.value":"Not used as raw training material or silently mined into constraints. Prompts, conversations, uploads, documents, tool input and private workspace context are processed for the requested task; an explicitly asked-for answer can yield only a derived candidate through the disclosed Spindle route, and Loom tells the user that it will train the governed constraint system.","trust-centre.statutory.training.field.feedback":"Optional feedback","trust-centre.statutory.training.field.feedback.value":"Separate and explicit opt-in; never required","trust-centre.statutory.training.field.feedbackReleaseUse":"Feedback included in Loom 1.0 training","trust-centre.statutory.training.field.feedbackReleaseUse.value":"No. Ordinary optional feedback is not training data, does not become a constraint and is not included in Loom 1.0. It can only guide future source-discovery and mining priorities, which still pass independent admission and validation. An explicitly asked-for answer is a separate, disclosed route for deriving a candidate constraint.","trust-centre.statutory.training.field.syntheticUse":"Use of synthetic data","trust-centre.statutory.training.field.syntheticUse.value":"Not for training. Synthetic material is used only for testing, evaluation and development, with its origin and generator retained in the test record.","trust-centre.statutory.training.field.syntheticModels":"Models used to create synthetic data","trust-centre.statutory.training.field.syntheticModels.value":"Not applicable to the training set. Test and development records identify the Dweve generator, solver, simulator or admitted external synthetic dataset used for that specific evaluation.","trust-centre.statutory.training.field.otherSourceUse":"Other source use and description","trust-centre.statutory.training.field.otherSourceUse.value":"No source bypasses the catalogue. Offline, self-digitised, commissioned or human-curated material must carry the same source identity, rights, purpose, evidence and admission decision before use.","trust-centre.statutory.training.field.signatory":"GPAI Code signatory","trust-centre.statutory.training.field.signatory.value":"Yes, Dweve is a signatory and Signatory Taskforce participant","trust-centre.statutory.training.field.rightsMeasures":"Rights-reservation measures","trust-centre.statutory.training.field.rightsMeasures.value":"Winnow source controls and Spindle rights, withdrawal and promotion state form the documented foundation","trust-centre.statutory.training.field.protocolInventory":"Complete honoured-protocol inventory","trust-centre.statutory.training.field.protocolInventory.value":"robots.txt, source-specific access and API terms, machine-readable and expressed rights reservations, per-domain policy, rate limits, authentication boundaries, rightsholder objections and removal decisions. Winnow publishes the acquisition implementation; Spindle records the resulting rights and withdrawal state.","trust-centre.statutory.training.field.illegalMeasures":"Illegal-content measures","trust-centre.statutory.training.field.illegalMeasures.value":"Deny unsafe and unauthorised targets; reject missing or incompatible rights; quarantine unresolved third-party payloads and privacy/confidentiality failures; validate claims against multiple sources before constraint extraction; preserve evidence and withdrawal state; and provide a working report, correction and removal route for rightsholders and affected people.","trust-centre.statutory.training.field.governedRoute":"Other processing information","trust-centre.statutory.training.field.governedRoute.value":"Winnow acquires, Spindle governs provenance and release state, and Loom consumes the versioned selection","trust-centre.statutory.training.field.updateCadence":"Update schedule","trust-centre.statutory.training.field.updateCadence.value":"At least every six months while further training content is admitted, or sooner when new content materially changes this summary; the update is published with the model state it covers.","trust-centre.loom.record.eyebrow":"Definitive system record","trust-centre.loom.record.title":"Dweve Loom in full","trust-centre.loom.record.body":"Loom is not one universal language model. It is a sparse cognitive system that plans a task-specific graph across learned models, explicit knowledge, memory, retrieval, reasoning, solvers, specialists, and verification. This record states what each part does, how the route is selected, what the output proves, and what still awaits release evidence.","trust-centre.loom.record.stat.experts":"Top-level domain specialists","trust-centre.loom.record.stat.experts.body":"Independent domain specialists with deeper specialist components, knowledge, constraints, and evaluations","trust-centre.loom.record.stat.active":"Selected on an average request","trust-centre.loom.record.stat.active.body":"An observed average, never a hard minimum or maximum","trust-centre.loom.record.stat.clusters":"Specialist clusters","trust-centre.loom.record.stat.clusters.body":"Foundational and specialised domains remain independently extensible","trust-centre.loom.record.stat.constraints":"Explicit constraint state","trust-centre.loom.record.stat.constraints.body":"Approximately 1.5 trillion across the catalogue; about 3 billion per domain specialist","trust-centre.loom.record.route.eyebrow":"Request execution","trust-centre.loom.record.route.title":"The task compiles its own cognitive graph","trust-centre.loom.record.route.body":"Only the threads that can contribute are activated. Each node has a typed responsibility, version, input, output, and evidence surface.","trust-centre.loom.record.route.read":"Read the task","trust-centre.loom.record.route.read.body":"Identify structure, modalities, domains, authority, and the form of answer required","trust-centre.loom.record.route.plan":"Plan the weave","trust-centre.loom.record.route.plan.body":"Choose perception, memory, retrieval, reasoning, solver, specialist, and verification responsibilities","trust-centre.loom.record.route.route":"Narrow the catalogue","trust-centre.loom.record.route.route.body":"Use staged signatures, neighbour search, gates, and structural similarity to select a bounded frontier","trust-centre.loom.record.route.solve":"Run native methods","trust-centre.loom.record.route.solve.body":"Use learned models where appropriate and explicit graphs, constraints, proofs, or solvers where the problem requires them","trust-centre.loom.record.route.verify":"Check the result","trust-centre.loom.record.route.verify.body":"Compare evidence, constraints, solver status, disagreement, and the declared completion contract","trust-centre.loom.record.route.express":"Express or refuse","trust-centre.loom.record.route.express.body":"Render the supported result coherently or return a typed refusal when no defensible path exists","trust-centre.loom.record.language.eyebrow":"Language boundary","trust-centre.loom.record.language.title":"Language presents the result, it does not decide it","trust-centre.loom.record.language.body":"Constraint learning, explicit graphs, solver outcomes, specialist work, and verification determine the supported result. Small bipolar language components render that completed result graph as coherent language; they do not decide it.","trust-centre.loom.record.systems.eyebrow":"Cognitive systems","trust-centre.loom.record.systems.title":"Different jobs keep their own machinery","trust-centre.loom.record.system.perception":"Multimodal perception","trust-centre.loom.record.system.perception.body":"Text, documents, images, audio, video, structured records, and sensor signals become typed spans, regions, events, entities, fields, and measurements","trust-centre.loom.record.system.memory":"Purpose-built memory","trust-centre.loom.record.system.memory.body":"Working, episodic, semantic, and procedural memory remain separate, policy-bounded, capturable responsibilities","trust-centre.loom.record.system.search":"Retrieval and ranking","trust-centre.loom.record.system.search.body":"Binary embeddings, retrieval, relevance, ranking, reranking, provenance, and policy are distinct stages rather than one opaque search answer","trust-centre.loom.record.system.learned":"Learned threads","trust-centre.loom.record.system.learned.body":"Fully bipolar transformers, micro models, embeddings, routers, and domain-specialist models use native one-bit, integer, or fixed-point machinery","trust-centre.loom.record.system.reasoning":"Eleven reasoning forms","trust-centre.loom.record.system.reasoning.body":"Deductive, inductive, abductive, analogical, causal, temporal, spatial, probabilistic, decision, metacognitive, and counterfactual reasoning","trust-centre.loom.record.system.solvers":"Solver fabric","trust-centre.loom.record.system.solvers.body":"Logical, constraint, optimisation, theorem, temporal, spatial, graph, automata, equilibrium, probabilistic, numerical, and synthesis families","trust-centre.loom.record.system.experts":"Domain specialists","trust-centre.loom.record.system.experts.body":"Each specialist carries its own knowledge graph, constraint set, source snapshot, version, and evaluation suite and can be upgraded independently","trust-centre.loom.record.system.verification":"Verification and refusal","trust-centre.loom.record.system.verification.body":"Checks evidence, contradictions, constraints, solver results, and route completion; unsupported work can end as an explicit refusal or unknown","trust-centre.loom.record.access.eyebrow":"Distribution","trust-centre.loom.record.access.title":"One model across five access routes","trust-centre.loom.record.access.fabric":"Fabric web","trust-centre.loom.record.access.fabric.body":"Human workspace route, planned for invite-only external access","trust-centre.loom.record.access.aura":"Aura","trust-centre.loom.record.access.aura.body":"Individual development route around the task and selected models","trust-centre.loom.record.access.api":"API","trust-centre.loom.record.access.api.body":"Controlled integration route with downstream documentation","trust-centre.loom.record.access.local":"Licensed local","trust-centre.loom.record.access.local.body":"Customer-controlled and air-gapped operation under the captured deterministic contract","trust-centre.loom.record.access.mesh":"Mesh","trust-centre.loom.record.access.mesh.body":"Distributed EU compute and deployment route","trust-centre.loom.record.open.eyebrow":"Open release evidence","trust-centre.loom.record.open.title":"What depends on a future event","trust-centre.loom.record.open.release":"Release identity","trust-centre.loom.record.open.release.body":"No external placement has occurred as of 1 August 2026; invite-only access is planned for 1 September 2026","trust-centre.loom.record.open.evaluations":"Measured evaluation results","trust-centre.loom.record.open.evaluations.body":"Methods and acceptance rules are public; the first external-release result and reviewer decision do not yet exist because that release has not occurred","trust-centre.loom.record.open.training":"Release-matched source admission","trust-centre.loom.record.open.training.body":"The 772 source families and admission rules are public; exact admitted-item, version, rights and provenance records are maintained in Spindle","trust-centre.loom.record.open.environment":"Compute and energy","trust-centre.loom.record.open.environment.body":"Hardware, historical training period, compute-node-hour and energy estimates, method, range and uncertainty are public; future specialist and release work will be metered prospectively","trust-centre.loom.record.open.article50":"Product output marking","trust-centre.loom.record.open.article50.body":"The disclosure, marking, trace and persistence contract is public; every route must bind pass evidence to its actual product version before external opening","trust-centre.loom.record.footer":"Canonical model detail · human-readable and machine-readable","trust-centre.training.method.eyebrow":"How Dweve trains","trust-centre.training.method.heading":"We extract knowledge|||before a model speaks","trust-centre.training.method.body":"Dweve does not train one giant language model to imitate its source material. It extracts explicit constraints and relationships, validates each against the source evidence, and keeps that link. This is constraint learning.","trust-centre.training.method.pills":"Admit|||Verify|||Reason","trust-centre.training.method.chrome":"Training boundary|||From source to supported answer|||Six stages|||Language is the final renderer","trust-centre.training.constraintLed":"Constraint-led knowledge|||The primary output is inspectable rules and relationships, not one opaque parameter mass","trust-centre.training.sourceBound":"Evidence per rule|||Every admitted constraint remains linked to source, licence, processing, and release lineage","trust-centre.training.languageBoundary":"Small language components|||Transformer-based language threads make a completed result coherent but do not determine it","trust-centre.training.admit":"Admit|||Resolve licence, commercial use, derivatives, provenance, attribution, third-party material, privacy, and confidentiality","trust-centre.training.extract":"Extract|||Turn admitted material into explicit constraints, entities, relations, examples, and typed knowledge","trust-centre.training.validate":"Validate|||Check each extracted rule or relationship against the identified source evidence","trust-centre.training.preserve":"Preserve|||Store evidence, source version, rights, lineage, conflicts, and withdrawal state with the knowledge","trust-centre.training.reason":"Reason|||Combine the relevant constraints, specialists, memory, reasoning modes, and solvers for the task","trust-centre.training.render":"Render|||Translate the verified result graph into coherent language without changing its supported meaning","trust-centre.training.inventory.eyebrow":"Supplied training-data inventory","trust-centre.training.inventory.title":"Every source family stays accountable","trust-centre.training.inventory.body":"Dweve has supplied most of its training and knowledge-source catalogue. This register publishes all 772 entries with their admission class, licence gate, permitted slice, and purpose. Listing is not blanket admission: every exact release and item still has to pass the deterministic gate and bind its evidence to Loom 1.0.","trust-centre.training.inventory.groups":"Source groups","trust-centre.training.inventory.sources":"Supplied entries","trust-centre.training.inventory.admitted":"Release-admitted entries","trust-centre.training.inventory.entries":"entries","trust-centre.training.inventory.sources.short":"sources","trust-centre.training.inventory.expand":"Open register","trust-centre.training.inventory.class.A0.title":"Core-clean","trust-centre.training.inventory.class.A0.body":"Public domain, CC0, or unconditional permissive licences with provenance","trust-centre.training.inventory.class.A1.title":"Attribution-clean","trust-centre.training.inventory.class.A1.body":"Attribution-based open sources admitted with an attribution ledger","trust-centre.training.inventory.class.A2.title":"Reciprocal-open","trust-centre.training.inventory.class.A2.body":"Share-alike or reciprocal sources isolated with their complete lineage","trust-centre.training.inventory.class.F.title":"Filterable family","trust-centre.training.inventory.class.F.body":"Only item-level records that resolve to an approved licence can pass","trust-centre.training.inventory.class.R.title":"Rights instrument","trust-centre.training.inventory.class.R.body":"A named registration, agreement, API term, territory, or commercial instrument must be archived","trust-centre.training.inventory.rule.eyebrow":"Deterministic admission","trust-centre.training.inventory.rule.title":"Every critical gate must pass","trust-centre.training.inventory.rule.body":"Catalogue, effective licence, commercial use, training and derivatives, provenance, attribution, third-party payload, privacy, and confidentiality are all required. A critical unknown cannot become an allowance.","trust-centre.training.inventory.column.source":"Source","trust-centre.training.inventory.column.class":"Class","trust-centre.training.inventory.column.gate":"Licence or gate","trust-centre.training.inventory.column.slice":"Usable slice","trust-centre.training.inventory.column.value":"Main value","trust-centre.training.inventory.column.status":"Release status","trust-centre.training.inventory.status.A0":"Directly admissible when provenance is captured","trust-centre.training.inventory.status.A1":"Attribution ledger required and retained","trust-centre.training.inventory.status.A2":"Reciprocal lineage isolated and retained","trust-centre.training.inventory.status.F":"Record-level licence filter applied before use","trust-centre.training.inventory.status.R":"Named rights instrument archived before use","trust-centre.training.inventory.footer":"Current source catalogue · exact evidence remains in the controlled Spindle ledger","trust-centre.training.inventory.policy":"Admission policy","trust-centre.loom.record.open.state":"Captured-state manifest schema","trust-centre.loom.record.open.state.body":"The public schema binds specialists, constraints, sources, configuration, routing, memory, tools, solvers, policies and evidence; an actual capture receives its own manifest","trust-centre.loom.record.state.value":"Adaptive · constraint-driven","trust-centre.loom.record.license.label":"Model licence","trust-centre.loom.record.license.value":"Proprietary","trust-centre.loom.record.license.body":"Public evidence and source licences do not grant rights to Loom, model weights, Dweve source code, operational packages, derivatives, or modification. Only separately identified Dweve libraries and tools are open source under their own licences.","trust-centre.nav.group.foundation":"Trust record","trust-centre.nav.group.model":"Model and data","trust-centre.nav.group.oversight":"Safety and oversight","trust-centre.nav.group.rights":"Rights and access","trust-centre.nav.choose":"Choose a Trust record","trust-centre.page.terms.crumb":"Commercial terms","trust-centre.page.terms.hero.eyebrow":"One commercial boundary","trust-centre.page.terms.hero.title":"Know what you receive|||before you commit","trust-centre.page.terms.hero.lede":"Fabric subscriptions, voluntary compute sharing, and Dweve licences are different arrangements. This record states what each includes, what remains yours, what remains Dweve's, and which document controls when terms meet.","trust-centre.page.terms.hero.actions":"Read the complete record|||Read the master terms","trust-centre.page.terms.glyph":"Commercial boundary|||Version 1.0 · pre-release","trust-centre.terms.version":"Content version 1.0|||The first public version takes effect on 1 September 2026","trust-centre.terms.acceptance":"Acceptance required|||Terms bind only through an applicable acceptance, order, or agreement","trust-centre.terms.scope":"Four connected records|||Master rules, Fabric, compute sharing, and licensing are separated","trust-centre.terms.map.eyebrow":"Document map","trust-centre.terms.map.heading":"One agreement stack|||without hidden overlap","trust-centre.terms.map.body":"The master terms provide the common legal frame. Product-specific records add only the terms needed for Fabric subscriptions, voluntary compute sharing, or licensed deployments. Mandatory law and a signed order remain visible in the hierarchy.","trust-centre.terms.map.pills":"Master|||Supplement|||Order","trust-centre.terms.map.chrome":"Document hierarchy|||Which record controls what|||Six records|||Specific terms do not silently rewrite another product","trust-centre.terms.master":"Common legal frame|||Identity, acceptance, responsibility, changes, disputes, and general service rules","trust-centre.terms.supplements":"Product-specific supplements|||Fabric, compute sharing, and licensing each state their own grant and limits","trust-centre.terms.mandatoryLaw":"Mandatory law preserved|||Consumer and statutory rights cannot be contracted away where the law prevents it","trust-centre.terms.fabricRecord":"Fabric subscription terms|||Named-user access, plan scope, data boundary, billing, and cancellation","trust-centre.terms.computeRecord":"Compute-sharing terms|||Voluntary contribution, device control, metering, settlement, and exit","trust-centre.terms.licensingRecord":"Licensing terms|||Deployment estate, proprietary rights, updates, restrictions, and source-right add-ons","trust-centre.terms.acceptableRecord":"Acceptable-use policy|||The authority, rights, and safety boundary applies to every access route","trust-centre.terms.privacyRecord":"Privacy policy|||Personal-data roles, rights, retention, and contact routes","trust-centre.terms.masterRecord":"Master terms of service|||The common legal agreement behind applicable product supplements","trust-centre.terms.lifecycle.eyebrow":"Version lifecycle","trust-centre.terms.lifecycle.heading":"Version 1.0 stays version 1.0|||until the first public launch","trust-centre.terms.lifecycle.body":"Pre-launch editing improves the same first version; it does not invent a public history nobody could have accepted. After launch, material changes receive a new content version, a dated change record, notice where required, and an archived predecessor.","trust-centre.terms.lifecycle.pills":"Publish|||Accept|||Archive","trust-centre.terms.lifecycle.chrome":"Terms lifecycle|||From pre-release to an accepted record|||Four stages|||No fictional version history","trust-centre.terms.prelaunch":"Pre-release now|||Version 1.0 is being completed before public availability","trust-centre.terms.notice":"Notice after launch|||Material changes are communicated through the applicable service and contact route","trust-centre.terms.history":"Public history only|||An archived predecessor appears only after an earlier public version actually existed","trust-centre.terms.publish":"Publish 1.0|||Freeze the effective text, machine records, date, and integrity manifest","trust-centre.terms.accept":"Accept in context|||Checkout, account creation, opt-in, or a signed order identifies the applicable terms","trust-centre.terms.operate":"Operate to the record|||Service delivery, metering, notices, support, and enforcement follow the accepted scope","trust-centre.terms.change":"Change transparently|||Version, explain, notify, and archive a material post-launch change","trust-centre.terms.record.eyebrow":"Canonical commercial record","trust-centre.terms.record.title":"The agreement stack in full","trust-centre.terms.record.body":"These pre-release terms separate common service rules, Fabric access, voluntary compute sharing and proprietary deployment licences. Each clause names the grant, boundary and exit without silently expanding another arrangement.","trust-centre.terms.record.version":"Content version","trust-centre.terms.record.published":"Record date","trust-centre.terms.record.effective":"Planned effective date","trust-centre.terms.record.status":"Record status","trust-centre.terms.record.status.value":"Version 1.0 · pre-release","trust-centre.terms.record.binding.title":"When these terms bind","trust-centre.terms.record.binding.body":"This page is the canonical pre-release record. A contract arises only through the acceptance shown at account creation, checkout, an explicit compute opt-in or a signed order. A signed order controls for its stated scope; product terms control their product; the master terms fill the common layer; mandatory law always prevails.","trust-centre.terms.record.master.eyebrow":"Common layer","trust-centre.terms.record.master.title":"Master service terms","trust-centre.terms.record.master.body":"Rules shared by every applicable Dweve service or licence. A product supplement or signed order adds specificity without silently changing another product.","trust-centre.terms.record.master.acceptance.title":"Acceptance is contextual","trust-centre.terms.record.master.acceptance.body":"The interface or order identifies the legal entity, account holder, service, price, term and documents being accepted. Browsing the Trust Centre alone is not acceptance.","trust-centre.terms.record.master.hierarchy.title":"A visible document hierarchy","trust-centre.terms.record.master.hierarchy.body":"Mandatory law comes first, followed by a signed order for its scope, the applicable product terms, the master terms and referenced policies. A more specific clause controls only the subject it addresses.","trust-centre.terms.record.master.version.title":"Version 1.0 until launch","trust-centre.terms.record.master.version.body":"Pre-launch edits remain content version 1.0 because no earlier public text was accepted. Material post-launch changes receive a new version, date, explanation, notice where required and an archived predecessor.","trust-centre.terms.record.master.mandatory.title":"Mandatory rights remain","trust-centre.terms.record.master.mandatory.body":"Nothing excludes consumer, privacy, product-safety, withdrawal, warranty or other statutory rights where applicable law does not permit that exclusion.","trust-centre.terms.record.master.contact.title":"One accountable provider","trust-centre.terms.record.master.contact.body":"Dweve B.V. in Arnhem, the Netherlands, is the contracting provider unless a signed order says otherwise. Legal and contractual questions use legal@dweve.com; ordinary service questions use the published support route.","trust-centre.terms.record.fabric.eyebrow":"Personal service","trust-centre.terms.record.fabric.title":"Fabric subscription terms","trust-centre.terms.record.fabric.body":"Fabric is the consumer-facing web and app service. Personal and Friends & Family access are named-user subscriptions; Aura, direct API access and Agent SDK rights are separate.","trust-centre.terms.record.fabric.scope.title":"The selected plan defines access","trust-centre.terms.record.fabric.scope.body":"Fabric Personal covers one named person. Fabric Friends & Family covers up to five named people with separate access. The order or checkout states the current monthly price and any uninterrupted-contract pre-order rate.","trust-centre.terms.record.fabric.accounts.title":"Accounts are personal","trust-centre.terms.record.fabric.accounts.body":"Each named person uses their own account and keeps credentials secure. Friends & Family administration can manage membership but does not merge personal workspaces or silently grant access to another member's content.","trust-centre.terms.record.fabric.data.title":"Your material is not raw training data","trust-centre.terms.record.fabric.data.body":"Prompts, conversations, uploads, documents, tool inputs and private workspace context are processed for the requested task, not used as raw training material and not silently mined into constraints. If Loom explicitly asks a knowledge-gap question, it tells the user that the answer will train the governed constraint system; only a derived, validated constraint may enter Spindle. Dweve does not claim ownership of user material or generated work merely because Fabric processed it.","trust-centre.terms.record.fabric.discovery.title":"External sources stay separate","trust-centre.terms.record.fabric.discovery.body":"A requested search may reveal an external source. Only that independently fetched source can enter ordinary governed candidate review. An explicitly asked-for user answer follows the separately disclosed derived-candidate route. When either route is used, Dweve identifies it in the same response.","trust-centre.terms.record.fabric.feedback.title":"Feedback is optional","trust-centre.terms.record.fabric.feedback.body":"A feedback request states its purpose. Ordinary feedback can guide future research and source-mining priorities but is not training data, does not become a constraint and is never required to keep using the service. An explicit knowledge-gap question is a separate route and tells the user that the answer will train the governed constraint system.","trust-centre.terms.record.fabric.billing.title":"Monthly billing is explicit","trust-centre.terms.record.fabric.billing.body":"The accepted checkout states price, tax, billing date, payment method and any pre-order condition. A recurring subscription renews monthly until cancelled; Dweve does not add a separate API or Aura entitlement unless the order says so.","trust-centre.terms.record.fabric.cancellation.title":"Cancel without hiding the exit","trust-centre.terms.record.fabric.cancellation.body":"The account holder can cancel future renewal through the account route or support. Access normally continues to the paid period's end, subject to mandatory withdrawal, refund and consumer rights.","trust-centre.terms.record.fabric.responsibility.title":"Assistance is not delegated authority","trust-centre.terms.record.fabric.responsibility.body":"Users remain responsible for the authority to provide material, the context in which output is used and any decision made from it. Dweve remains responsible for operating Fabric within the accepted service, privacy and safety boundaries.","trust-centre.terms.record.compute.eyebrow":"Separate opt-in","trust-centre.terms.record.compute.title":"Voluntary compute-sharing terms","trust-centre.terms.record.compute.body":"Compute sharing is an optional Personal route, not a Friends & Family substitute, investment, employment arrangement, token scheme or public compute market.","trust-centre.terms.record.compute.eligibility.title":"Eligibility follows a paid Personal month","trust-centre.terms.record.compute.eligibility.body":"A Fabric Personal subscriber can become eligible after the first paid month and after the device, region, security and availability checks shown in the opt-in flow. Friends & Family membership alone does not create eligibility.","trust-centre.terms.record.compute.voluntary.title":"Contribution is always voluntary","trust-centre.terms.record.compute.voluntary.body":"The contributor must opt in separately and can pause or revoke contribution. Fabric use is not conditioned on sharing compute, and Dweve cannot use a device outside the accepted availability and resource boundary.","trust-centre.terms.record.compute.control.title":"The owner keeps device control","trust-centre.terms.record.compute.control.body":"The contributor sets supported schedules and limits. The client exposes current state, accepted work, resource use and stop controls; revocation prevents new work and safely concludes or expires accepted work.","trust-centre.terms.record.compute.workloads.title":"Sandboxed Dweve work only","trust-centre.terms.record.compute.workloads.body":"Only signed, policy-scoped Dweve workloads may run. Customer prompts, private documents and unrelated customer data are not placed on contributor devices. Selvedge-style isolation, capability limits and receipts define the execution boundary.","trust-centre.terms.record.compute.measurement.title":"Measurement produces a receipt","trust-centre.terms.record.compute.measurement.body":"Accepted work, availability, resource class, useful completion, integrity checks and settlement status are metered under a versioned method. Failed, rejected, unverifiable or policy-violating work does not earn settlement.","trust-centre.terms.record.compute.settlement.title":"Settlement covers Personal under the stated rules","trust-centre.terms.record.compute.settlement.body":"Qualifying contribution can settle the contributor's Fabric Personal subscription under the current programme rules. It is internal service accounting, not cash, cryptocurrency, a transferable credit or a promise of profit.","trust-centre.terms.record.compute.costs.title":"Device costs stay visible","trust-centre.terms.record.compute.costs.body":"The contributor remains responsible for electricity, connectivity, equipment wear, tax and other local costs. The opt-in must show enough metering information to decide whether contribution remains worthwhile.","trust-centre.terms.record.compute.integrity.title":"Tampering closes the route","trust-centre.terms.record.compute.integrity.body":"Dweve may reject work, suspend contribution or require re-verification where software, receipts, workload isolation, device identity or measured results are altered, replayed or cannot be verified.","trust-centre.terms.record.compute.exit.title":"Exit returns to ordinary paid access","trust-centre.terms.record.compute.exit.body":"Stopping compute sharing does not delete the Fabric account. The user can continue or renew Personal through the ordinary paid route, cancel the subscription separately or exercise applicable consumer rights.","trust-centre.terms.record.licensing.eyebrow":"Organisation deployment","trust-centre.terms.record.licensing.title":"Proprietary licensing terms","trust-centre.terms.record.licensing.body":"Dweve models, products and commercial runtime remain proprietary. A licence grants defined operational use; it does not quietly transfer the model, source code or intellectual property.","trust-centre.terms.record.licensing.grant.title":"A scoped right to operate","trust-centre.terms.record.licensing.grant.body":"The signed order identifies the customer, products, environments, term, geography, users, support and permitted purpose. Rights not expressly granted remain with Dweve or the applicable third-party rightsholder.","trust-centre.terms.record.licensing.estates.title":"The order defines the estate","trust-centre.terms.record.licensing.estates.body":"Solution, Platform, Enterprise and Global arrangements differ by the deployment estate stated in the order. Licensed local or air-gapped operation may be included; no per-query fee applies inside that estate unless the order expressly states one.","trust-centre.terms.record.licensing.proprietary.title":"Loom and the commercial stack stay proprietary","trust-centre.terms.record.licensing.proprietary.body":"The licence does not open-source Loom, its learned components, constraint state, model weights, Dweve products, Kera, operational packages or confidential implementation material.","trust-centre.terms.record.licensing.openSource.title":"Open foundations keep their own licences","trust-centre.terms.record.licensing.openSource.body":"Separately identified Dweve repositories published under Apache 2.0 remain governed by that repository's licence. Their presence does not convert the proprietary platform or model into open source.","trust-centre.terms.record.licensing.sourceRights.title":"Deeper source rights are a separate add-on","trust-centre.terms.record.licensing.sourceRights.body":"Escrow, secured review, operational source delivery, adaptation or independence rights exist only when a signed order names that option, its materials, triggers, confidentiality and continuing-use boundary.","trust-centre.terms.record.licensing.operation.title":"Customer control and Dweve responsibility are separated","trust-centre.terms.record.licensing.operation.body":"The customer controls authorised deployment, users, inputs, local infrastructure and use context. Dweve supplies the licensed software, documentation, applicable updates and support stated in the order.","trust-centre.terms.record.licensing.updates.title":"Updates do not erase a captured contract","trust-centre.terms.record.licensing.updates.body":"Adaptive service operation and new constraints are documented separately from sealed local packages. Supported updates, compatibility, captured-state manifests and end-of-support obligations follow the order and release record.","trust-centre.terms.record.licensing.restrictions.title":"No redistribution or competing service","trust-centre.terms.record.licensing.restrictions.body":"Unless expressly granted, the customer may not redistribute the proprietary stack, expose it as an external service, bypass access controls, extract protected implementation material or use it to build a competing general-purpose AI model.","trust-centre.terms.record.licensing.termination.title":"Termination respects retained rights","trust-centre.terms.record.licensing.termination.body":"On expiry or termination, operational rights end except for any survival, export, transition, escrow or continuing-use rights expressly stated in the order. Customer data and customer-owned work remain the customer's.","trust-centre.terms.record.downloads":"Machine-readable terms","trust-centre.terms.record.download.index":"Commercial terms index","trust-centre.terms.record.download.fabric":"Fabric terms","trust-centre.terms.record.download.compute":"Compute-sharing terms","trust-centre.terms.record.download.licensing":"Licensing terms","trust-centre.generated.marking.eyebrow":"Article 50 marking","trust-centre.generated.marking.heading":"One release contract|||for every output surface","trust-centre.generated.marking.body":"The marking policy covers Loom text and code output plus the images Dweve ships. C2PA Content Credentials carry the signed, machine-readable provenance record.","trust-centre.generated.marking.pills":"Scope|||C2PA|||Detection","trust-centre.generated.marking.chrome":"Marking policy|||Modality matrix|||Text, code, and image output|||Video excluded","trust-centre.generated.marking.scope":"Provider scope|||Text, code, and image output are covered from the route's first external access; Loom does not generate video","trust-centre.generated.marking.machine":"Machine-readable layer|||C2PA Content Credentials carry the signer, generation action, timestamp and asset identifier","trust-centre.generated.marking.release":"Release condition|||The route stays closed until marking, detection, persistence and integrity checks pass for its real version","trust-centre.generated.marking.text":"Text|||C2PA Content Credentials are the machine-readable and detectable provenance layer for text exports","trust-centre.generated.marking.audio":"Audio output|||No generic audio route is claimed in this release-bound record","trust-centre.generated.marking.image":"Image output|||C2PA Content Credentials carry the signed, machine-readable provenance record","trust-centre.generated.marking.video":"Video generation|||Loom 1.0 does not generate video, so no video marking surface is claimed","trust-centre.generated.verification.eyebrow":"Detection path","trust-centre.generated.verification.heading":"Verification starts with the file|||not a guess about the text","trust-centre.generated.verification.body":"A C2PA-compliant validator reads the signed Content Credential on an original export. The record is a provenance signal, not proof of truth or authorship.","trust-centre.generated.verification.pills":"Keep the export|||Inspect the manifest|||Check the route","trust-centre.generated.verification.chrome":"Verification path|||How to inspect an output|||Four steps|||No cross-provider detector","trust-centre.generated.verification.validator":"Public validator|||Use the open-source C2PA tool, Adobe Content Credentials inspector or another C2PA-aware application","trust-centre.generated.verification.manifest":"Signed record|||Inspect the Dweve organisation, generation action, timestamp, asset identifier and manifest status","trust-centre.generated.verification.audio":"Audio supplement|||No additional audio detector is claimed in this release-bound record","trust-centre.generated.verification.keep":"Keep|||Retain the original file or export; a screenshot or metadata-cleaned copy has no recoverable embedded binding","trust-centre.generated.verification.inspect":"Inspect|||Read the signed organisation, action, timestamp, identifier, ingredients and later edits","trust-centre.generated.verification.compare":"Compare|||Check the route manifest and the declared format or compression limits","trust-centre.generated.verification.request":"Request|||Ask for a scoped record when a route-specific investigation needs more evidence","trust-centre.generated.manifest.eyebrow":"Per-release manifest","trust-centre.generated.manifest.heading":"Every release names the surface|||and the evidence behind its mark","trust-centre.generated.manifest.body":"The registry carries a marking object beside Loom's identity, architecture and licence. It records the policy version, route surfaces, assertions, detector, round-trip evidence and the integrity record used to verify the public files.","trust-centre.generated.manifest.pills":"Versioned|||Surface-specific|||Hash-linked","trust-centre.generated.manifest.chrome":"Release evidence|||Marking manifest|||Three obligations|||Route-readiness evidence","trust-centre.generated.manifest.policy":"Policy version|||The route identifies the marking policy and the exact Loom release it implements","trust-centre.generated.manifest.surfaces":"Surface contract|||API, Fabric web, Aura's coding-agent route, licensed local deployment and Mesh are recorded separately","trust-centre.generated.manifest.evidence":"Pass evidence|||C2PA signing, validation, transformations, and route limitations travel with the release decision","trust-centre.generated.manifest.effective":"Effective date|||The route opening decision supplies the date; no external route is open yet","trust-centre.generated.manifest.integrity":"Integrity link|||The policy, registry and release-readiness record are included in the SHA-256 public record","trust-centre.close.titleLead":"A record should be challengeable","trust-centre.close.titleAccent":"or it is only a claim","trust-centre.close.secondary":"Open the record map","trust-centre.close.panel.title":"Using the record","trust-centre.close.panel.tag":"Three habits","trust-centre.close.panel.row.0.label":"Read the status first","trust-centre.close.panel.row.0.value":"Every record states whether it is verified, prepared or pending, and a missing field stays visible rather than being filled in.","trust-centre.close.panel.row.1.label":"Take the machine copy","trust-centre.close.panel.row.1.value":"The same statements are published as JSON alongside the pages, so a reader can check them without reading prose.","trust-centre.close.panel.row.2.label":"Contest what is wrong","trust-centre.close.panel.row.2.value":"Rights, safety and information requests go through one recorded route, with the subject chosen at the desk.","trust-centre.close.panel.foot":"Public record, updated 1 August 2026"}