RAINavigator

Append-only, hash-chained audit vault (WORM object-lock storage, AES-256 at rest, TLS 1.3 in transit). Guarantees tamper-evidence within the organization's trust domain — which stops your own team, but not an admin who can rebuild the vault. Pair with an external trust anchor and key ceremonies outside the operating team for evidence that holds against the insider scenario.

• article obligationArt. 12 — Record-Keeping / Logging → WORM / Immutable Audit VaultArt. 12 — Record-Keeping / Logging
• article obligationArt. 26 — Deployer Obligations → WORM / Immutable Audit VaultArt. 26 — Deployer Obligations
• control objectiveArt. 12 — Record-Keeping / Logging → CO: Log Completeness & Coverage → WORM / Immutable Audit VaultArt. 12 — Record-Keeping / LoggingCO: Log Completeness & Coverage
• control objectiveArt. 12 — Record-Keeping / Logging → CO: Log Integrity & Non-Repudiation → WORM / Immutable Audit VaultArt. 12 — Record-Keeping / LoggingCO: Log Integrity & Non-Repudiation
• evidence artefactArt. 12 — Record-Keeping / Logging → Event Logs & Decision Traces → produced by WORM / Immutable Audit VaultArt. 12 — Record-Keeping / LoggingEvent Logs & Decision Traces

HITL escalation queue & review UI ('Human-as-a-Tool': the agent calls the human like any other tool via propose-action objects). Confidence- and risk-threshold routing, SLA timers, structured accept/modify/reject verdicts with digital reviewer signature at gate release — each verdict is itself Art. 14 evidence and feeds the active-learning loop.

• article obligationArt. 86 — Right to explanation of individual decision-making → HITL Escalation Queue & Review UIArt. 86 — Right to explanation of individual decision-making
• article obligationArt. 14 — Human Oversight → HITL Escalation Queue & Review UIArt. 14 — Human Oversight
• article obligationArt. 26 — Deployer Obligations → HITL Escalation Queue & Review UIArt. 26 — Deployer Obligations
• control objectiveArt. 14 — Human Oversight → CO: Oversight Competence & Authority → HITL Escalation Queue & Review UIArt. 14 — Human OversightCO: Oversight Competence & Authority
• evidence artefactArt. 14 — Human Oversight → Human-Oversight Protocol & Intervention Records → produced by HITL Escalation Queue & Review UIArt. 14 — Human OversightHuman-Oversight Protocol & Intervention Records

Operator stop controls and degraded-mode fallbacks; real-time override (veto) channels for HOTL operation.

• article obligationArt. 14 — Human Oversight → Kill Switch / Graceful DegradationArt. 14 — Human Oversight

Computes a probabilistic confidence score for every output and holds the transaction when the score falls below the workflow's regulatory threshold.

• article obligationArt. 14 — Human Oversight → Confidence Scoring & Threshold GateArt. 14 — Human Oversight
• design patternArt. 14 — Human Oversight → Confidence-Threshold HITL Routing → Confidence Scoring & Threshold GateArt. 14 — Human OversightConfidence-Threshold HITL Routing

Feature attributions for classical ML, reasoning-trace summaries for GenAI — feeds the human reviewer and the technical file.

• article obligationArt. 86 — Right to explanation of individual decision-making → Explainability API (SHAP/LIME/CoT)Art. 86 — Right to explanation of individual decision-making
• article obligationArt. 13 — Transparency to Deployers → Explainability API (SHAP/LIME/CoT)Art. 13 — Transparency to Deployers
• article obligationArt. 14 — Human Oversight → Explainability API (SHAP/LIME/CoT)Art. 14 — Human Oversight

Representativeness checks, bias metrics and mitigation per ISO/IEC 5259; versioned datasets with lineage.

• article obligationArt. 10 — Data Governance → Bias Testing & Data Quality PipelineArt. 10 — Data Governance
• control objectiveArt. 15 — Accuracy, Robustness, Cybersecurity → CO: Adversarial Robustness Verified → Bias Testing & Data Quality PipelineArt. 15 — Accuracy, Robustness, CybersecurityCO: Adversarial Robustness Verified
• evidence artefactArt. 10 — Data Governance → Dataset Documentation & Bias Reports → produced by Bias Testing & Data Quality PipelineArt. 10 — Data GovernanceDataset Documentation & Bias Reports

Provenance tracking of datasets, features and embeddings; write-time attribution (source, actor, timestamp, confidence).

• article obligationArt. 10 — Data Governance → Data Lineage & VersioningArt. 10 — Data Governance
• control objectiveArt. 10 — Data Governance → CO: Data Quality & Origin Management → Data Lineage & VersioningArt. 10 — Data GovernanceCO: Data Quality & Origin Management
• evidence artefactArt. 10 — Data Governance → Dataset Documentation & Bias Reports → produced by Data Lineage & VersioningArt. 10 — Data GovernanceDataset Documentation & Bias Reports

Hierarchical trace spans for every sub-task, prompt, retrieved document and API call — the reconstructible decision path for Art. 12/14 and PLD disclosure.

• article obligationArt. 12 — Record-Keeping / Logging → OpenTelemetry / FCoT TracingArt. 12 — Record-Keeping / Logging
• control objectiveArt. 12 — Record-Keeping / Logging → CO: Log Completeness & Coverage → OpenTelemetry / FCoT TracingArt. 12 — Record-Keeping / LoggingCO: Log Completeness & Coverage
• control objectiveArt. 13 — Transparency to Deployers → CO: System Traceability & Decision Transparency → OpenTelemetry / FCoT TracingArt. 13 — Transparency to DeployersCO: System Traceability & Decision Transparency
• evidence artefactArt. 12 — Record-Keeping / Logging → Event Logs & Decision Traces → produced by OpenTelemetry / FCoT TracingArt. 12 — Record-Keeping / LoggingEvent Logs & Decision Traces

Escalation triggers built from two independent signals, because raw model confidence is uncalibrated: calibrated trust scores (prompt relevance, similarity to historic successes, cross-model consistency) plus deterministic risk scores (sensitive categories, transaction value, protected data) — either crossing its threshold forces human review.

• article obligationArt. 14 — Human Oversight → Trust & Risk Dual ScoringArt. 14 — Human Oversight

Faithfulness scoring of answers against retrieved sources; deterministic fallback instead of hallucination; schema-validated structured output.

• article obligationArt. 15 — Accuracy, Robustness, Cybersecurity → Output Rails / Groundedness CheckArt. 15 — Accuracy, Robustness, Cybersecurity
• control objectiveArt. 50 — Transparency Duties → CO: AI Interaction & Content Disclosure → Output Rails / Groundedness CheckArt. 50 — Transparency DutiesCO: AI Interaction & Content Disclosurepublication-gate verification
• evidence artefactArt. 15 — Accuracy, Robustness, Cybersecurity → Guardrail Telemetry & Sanitization Records → produced by Output Rails / Groundedness CheckArt. 15 — Accuracy, Robustness, CybersecurityGuardrail Telemetry & Sanitization Records

Automated detection, pseudonymisation and blocking of personal data in inputs, retrievals and outputs.

• article obligationArt. 10 — Data Governance → PII Scrubbing / DLP-NER LayerArt. 10 — Data Governance
• design patternArt. 10 — Data Governance → Inline PII/PHI Tokenisation → PII Scrubbing / DLP-NER LayerArt. 10 — Data GovernanceInline PII/PHI Tokenisation

Practice-derived component: converts feature attributions (SHAP or an equivalent attribution method) into an individually understandable, legally defensible explanation of an adverse decision — the role the AI system played, the main elements the decision rested on, and counterfactual scenarios stating what would have had to differ for a different outcome. Reason codes are generated from the decisioning path, not from a marketing template, and every issued letter is retained with the model version and the attribution run behind it. Honesty condition: a reason is only usable if acting on it would actually change the outcome, which non-monotonic feature interactions can break (see the post-hoc instability threat).

• article obligationArt. 86 — Right to explanation of individual decision-making → Adverse-Decision Reason GeneratorArt. 86 — Right to explanation of individual decision-making
• article obligationArt. 13 — Transparency to Deployers → Adverse-Decision Reason GeneratorArt. 13 — Transparency to Deployers

Relevance, freshness and per-user permission checks on every retrieved chunk; curated, versioned index.

• article obligationArt. 10 — Data Governance → Retrieval Rails (ACL-aware RAG)Art. 10 — Data Governance

Policy-as-code decision point (PDP) with enforcement points (PEP) in front of every tool call: versioned policies in Git, microsecond evaluation, typed action schemas — authorization decided outside the model's reasoning space, never in the prompt.

• article obligationArt. 9 — Risk Management → Deterministic Policy Engine (OPA / Cedar)Art. 9 — Risk Management

Abstraction layer decoupling application logic from model providers: dynamic routing on capability, cost, latency SLA and regulatory constraint (sensitive-data classes pinned to ZDR private/VPC endpoints or on-prem open-weight instances); real-time health monitoring with automatic fallback to secondary endpoints or local fine-tuned models on outage/latency spikes. Discharges resilience duties (DORA-class), prevents provider lock-in, and makes model deprecations a routing-table change instead of a re-architecture. Router decisions are logged into the decision trace — model version per event is an audit-packet field.

• design patternArt. 15 — Accuracy, Robustness, Cybersecurity → Model Abstraction & Graceful Fallback → Multi-Model Router & Fallback AbstractionArt. 15 — Accuracy, Robustness, CybersecurityModel Abstraction & Graceful Fallback

Pre-model validation of user input: injection detection, topic blocking, encoding checks.

• article obligationArt. 15 — Accuracy, Robustness, Cybersecurity → Input Rails / Prompt ShieldsArt. 15 — Accuracy, Robustness, Cybersecurity
• control objectiveArt. 15 — Accuracy, Robustness, Cybersecurity → CO: Runtime Injection Defense → Input Rails / Prompt ShieldsArt. 15 — Accuracy, Robustness, CybersecurityCO: Runtime Injection Defense
• evidence artefactArt. 15 — Accuracy, Robustness, Cybersecurity → Guardrail Telemetry & Sanitization Records → produced by Input Rails / Prompt ShieldsArt. 15 — Accuracy, Robustness, CybersecurityGuardrail Telemetry & Sanitization Records

Synthetic-content labelling & watermarking: visible disclosure plus machine-readable provenance (C2PA Content Credentials) embedded in generated images, audio and video; metadata identifying artificial origin survives common transformations. Discharges Art. 50(2)/(4) for deepfakes and synthetic media; verification telemetry (watermark presence/validity checks at publication gates) is the corresponding evidence stream.

• article obligationArt. 50 — Transparency Duties → Synthetic-Content Labelling / WatermarkingArt. 50 — Transparency Duties
• control objectiveArt. 50 — Transparency Duties → CO: AI Interaction & Content Disclosure → Synthetic-Content Labelling / WatermarkingArt. 50 — Transparency DutiesCO: AI Interaction & Content Disclosure

Cost/iteration caps, loop detection, anomaly-triggered mandatory approval (CodeBuddy 'suspicious command override').

• article obligationArt. 9 — Risk Management → Watchdog Supervisor & Rate LimitingArt. 9 — Risk Management
• article obligationArt. 15 — Accuracy, Robustness, Cybersecurity → Watchdog Supervisor & Rate LimitingArt. 15 — Accuracy, Robustness, Cybersecurity

Governed runtime workspace operationalizing Art. 10: traceable lineage for every RAG chunk and training record at execution time, canonical version-controlled business glossary (documents Art. 10(2)(d) baseline assumptions), and continuous data-quality monitoring with threshold alerts and logged remediation for the Art. 10(3) 'error-free and complete' standard.

• article obligationArt. 10 — Data Governance → Sovereign Context LayerArt. 10 — Data Governance
• evidence artefactArt. 10 — Data Governance → Dataset Documentation & Bias Reports → produced by Sovereign Context LayerArt. 10 — Data GovernanceDataset Documentation & Bias Reportslineage + quality monitoring records

Per-agent identities, short-lived scoped tokens, OBO flow enforcement — the identity substrate of agentic zero trust.

• control objectiveArt. 15 — Accuracy, Robustness, Cybersecurity → CO: Non-Human Identity Governance → Agent Identity & Access (IdP)Art. 15 — Accuracy, Robustness, CybersecurityCO: Non-Human Identity Governanceidentity hygiene is part of the cybersecurity duty for agentic systems

Agents never hold target-system keys; the gateway injects centrally managed credentials after policy checks.

• control objectiveArt. 12 — Record-Keeping / Logging → CO: Log Access & Retention Governance → Central Credential VaultArt. 12 — Record-Keeping / LoggingCO: Log Access & Retention Governance

Continuous evaluation against golden sets and sampled human verdicts; raises drift alerts and feeds the recertification cycle.

• control objectiveArt. 72/73 — Post-Market Monitoring & Incidents → CO: Performance Monitoring & Drift Management → Model Drift & Accuracy MonitorArt. 72/73 — Post-Market Monitoring & IncidentsCO: Performance Monitoring & Drift Management

AI register & model registry: central inventory of every model, agent, RAG pipeline and embedded third-party SaaS AI across the estate, with factsheets per asset. v2.0 duty: every application — internal, open-source or procured — continuously publishes a machine-readable AI-BOM and Factsheet into the register; an asset without a current AI-BOM is an inventory gap, not a formality. Feeds Colorado AIA/ LL144 disclosure duties and the Art. 11 technical file; the enforcement backstop is Shadow-AI discovery on the risk register.

• article obligationArt. 16 — Obligations of providers of high-risk AI systems → AI Register & Model Registry / FactsheetsArt. 16 — Obligations of providers of high-risk AI systems
• article obligationArt. 49 — Registration in the EU database → AI Register & Model Registry / FactsheetsArt. 49 — Registration in the EU database
• article obligationArt. 11 — Technical Documentation → AI Register & Model Registry / FactsheetsArt. 11 — Technical Documentation
• article obligationArt. 13 — Transparency to Deployers → AI Register & Model Registry / FactsheetsArt. 13 — Transparency to Deployers
• control objectiveArt. 9 — Risk Management → CO: Pre-Deployment AI Risk & Impact Assessment → AI Register & Model Registry / FactsheetsArt. 9 — Risk ManagementCO: Pre-Deployment AI Risk & Impact Assessment
• control objectiveArt. 26 — Deployer Obligations → CO: Embedded-AI Vendor Governance → AI Register & Model Registry / FactsheetsArt. 26 — Deployer ObligationsCO: Embedded-AI Vendor Governancethird-party SaaS AI belongs in the same inventory
• evidence artefactArt. 11 — Technical Documentation → Technical Documentation (Annex IV) → produced by AI Register & Model Registry / FactsheetsArt. 11 — Technical DocumentationTechnical Documentation (Annex IV)automated factsheet generation
• evidence artefactArt. 11 — Technical Documentation → AI Bill of Materials (AI-BOM) & Factsheets → produced by AI Register & Model Registry / FactsheetsArt. 11 — Technical DocumentationAI Bill of Materials (AI-BOM) & Factsheets
• evidence artefactArt. 43 — Conformity Assessment → Technical Documentation (Annex IV) → produced by AI Register & Model Registry / FactsheetsArt. 43 — Conformity AssessmentTechnical Documentation (Annex IV)automated factsheet generation

Scoring model: jurisdiction (CLOUD Act exposure), zero-data-retention, BYOK support, audit evidence (C5/AIC4/ISO 42001/EN 18286:2026), tenant isolation.

• article obligationArt. 53 — Obligations for providers of general-purpose AI models → Vendor & Model Due-Diligence KitArt. 53 — Obligations for providers of general-purpose AI models
• control objectiveArt. 26 — Deployer Obligations → CO: Embedded-AI Vendor Governance → Vendor & Model Due-Diligence KitArt. 26 — Deployer ObligationsCO: Embedded-AI Vendor Governance
• design patternArt. 25 — Value Chain / Role Flip → Zero-Data-Retention Vendor Binding → Vendor & Model Due-Diligence KitArt. 25 — Value Chain / Role FlipZero-Data-Retention Vendor Binding
• evidence artefactArt. 26 — Deployer Obligations → Vendor & Model Due-Diligence Records → produced by Vendor & Model Due-Diligence KitArt. 26 — Deployer ObligationsVendor & Model Due-Diligence Records

AMD SEV / Intel TDX: data protected from the cloud operator even in memory during inference.

• article obligationArt. 15 — Accuracy, Robustness, Cybersecurity → Confidential Computing EnclavesArt. 15 — Accuracy, Robustness, Cybersecurity

The engine behind inline tokenisation: pre-model interception that replaces identifiers with reversible tokens before a payload leaves the isolation boundary, plus a detokenisation gate that re-identifies only for authorised callers inside the boundary and logs every re-identification. Complements the DLP/NER scrubbing layer, which blocks or masks rather than preserving reversible reference.

• design patternArt. 10 — Data Governance → Inline PII/PHI Tokenisation → PII/PHI Redaction & Tokenisation EngineArt. 10 — Data GovernanceInline PII/PHI Tokenisation

Agent conversational and working memory managed as a books-and-records object rather than a cache: retention schedules per regime, immutability where records rules demand it, deletion paths for erasure requests that do not break the audit trail, and export in a form a supervisor can read. Where an agent's memory carries a business communication or a decision rationale, it is a record — the storage tier does not decide that.

• control objectiveArt. 12 — Record-Keeping / Logging → CO: Log Access & Retention Governance → Agent Memory Record StoreArt. 12 — Record-Keeping / LoggingCO: Log Access & Retention Governance
• evidence artefactArt. 12 — Record-Keeping / Logging → Event Logs & Decision Traces → produced by Agent Memory Record StoreArt. 12 — Record-Keeping / LoggingEvent Logs & Decision Traces

The disclosure surface at the engagement layer: an AI-interaction notice on every channel a natural person can reach (web, app, voice, chat, social, marketplace), machine-readable provenance marking on generated or manipulated output, and a disclosure record per interaction that can be produced on request. Sits at the interface, not in the model — a model-side label that the frontend drops is not a disclosure.

• article obligationArt. 50 — Transparency Duties → Interface Transparency & Content-Marking LayerArt. 50 — Transparency Dutiesthe disclosure duty lands at the interface, not in the model
• control objectiveArt. 50 — Transparency Duties → CO: AI Interaction & Content Disclosure → Interface Transparency & Content-Marking LayerArt. 50 — Transparency DutiesCO: AI Interaction & Content Disclosuredisclosure notice + provenance marking per surface
• evidence artefactArt. 13 — Transparency to Deployers → Instructions for Use / Transparency Docs → produced by Interface Transparency & Content-Marking LayerArt. 13 — Transparency to DeployersInstructions for Use / Transparency Docsdisclosure records per interaction surface
• evidence artefactArt. 50 — Transparency Duties → Instructions for Use / Transparency Docs → produced by Interface Transparency & Content-Marking LayerArt. 50 — Transparency DutiesInstructions for Use / Transparency Docsdisclosure records per interaction surface

Software bill of materials incl. model weights and datasets; automated vulnerability patching pipeline.

• article obligationArt. 53 — Obligations for providers of general-purpose AI models → SBOM & Dependency ManagementArt. 53 — Obligations for providers of general-purpose AI models
• evidence artefactArt. 11 — Technical Documentation → AI Bill of Materials (AI-BOM) & Factsheets → produced by SBOM & Dependency ManagementArt. 11 — Technical DocumentationAI Bill of Materials (AI-BOM) & Factsheetssoftware-dependency half of the manifest

Every agent action leaves through one interception point that validates the call against a declared schema and an allow-list of permitted effects, rejects out-of-contract arguments, and records the attempt whether it passed or not. Turns 'the agent may call the payment API' into 'the agent may call this method, with these fields, inside these limits'.

• control objectiveArt. 15 — Accuracy, Robustness, Cybersecurity → CO: Runtime Injection Defense → Tool-Use Boundary ProxyArt. 15 — Accuracy, Robustness, CybersecurityCO: Runtime Injection Defense
• evidence artefactArt. 15 — Accuracy, Robustness, Cybersecurity → Guardrail Telemetry & Sanitization Records → produced by Tool-Use Boundary ProxyArt. 15 — Accuracy, Robustness, CybersecurityGuardrail Telemetry & Sanitization Records

Issues ephemeral, per-task, narrowly scoped credentials to agents and revokes them on task completion, escalation or anomaly — the 'no standing credentials' principle in a component. Distinct from the credential vault (which holds long-lived secrets centrally): the broker's product is a credential that expires before it can be exfiltrated and reused.

• control objectiveArt. 15 — Accuracy, Robustness, Cybersecurity → CO: Non-Human Identity Governance → Non-Human Identity Credential BrokerArt. 15 — Accuracy, Robustness, CybersecurityCO: Non-Human Identity Governanceidentity hygiene is part of the cybersecurity duty for agentic systems

One procedure reconciling AI Act Art. 73, GDPR Art. 33 (72h), DORA and NIS2 (24h/72h) timelines and recipients.

• article obligationArt. 72/73 — Post-Market Monitoring & Incidents → Unified Incident-Response RunbookArt. 72/73 — Post-Market Monitoring & Incidents
• evidence artefactArt. 72/73 — Post-Market Monitoring & Incidents → Post-Market Monitoring Plan & Incident Reports → produced by Unified Incident-Response RunbookArt. 72/73 — Post-Market Monitoring & IncidentsPost-Market Monitoring Plan & Incident Reports

Practice-derived component: saliency, heatmap or region-proposal overlays rendered on the study itself, so the reviewing clinician can check the anatomical plausibility of the finding — whether the model looked where the pathology is — instead of accepting a score. Overlay stability across reconstructions is monitored, because an unstable overlay is a false assurance rather than an explanation.

• article obligationArt. 13 — Transparency to Deployers → Visual Explainability for Clinical ReviewArt. 13 — Transparency to DeployersVisual explainability for clinical review answers Art. 13 in an imaging context; conditioned for the same reason as the DICOM edges.
• article obligationArt. 14 — Human Oversight → Visual Explainability for Clinical ReviewArt. 14 — Human OversightSurfaced on CV screening in Study 6, where clinical review has no meaning.

Continuously updated risk register wired to runtime posture: threat-model deltas, open defects, control status, exposure per system. Includes Shadow-AI discovery — continuous scanning for unsanctioned agents, MCP servers and AI API usage outside the register; an unregistered agent is an unmanaged Art. 12/26 liability and the empirical driver of proportionate (not blanket) controls.

• article obligationArt. 16 — Obligations of providers of high-risk AI systems → Live Risk Register / Posture ManagementArt. 16 — Obligations of providers of high-risk AI systems
• article obligationArt. 9 — Risk Management → Live Risk Register / Posture ManagementArt. 9 — Risk Management
• evidence artefactArt. 9 — Risk Management → Risk Management File / Live Register → produced by Live Risk Register / Posture ManagementArt. 9 — Risk ManagementRisk Management File / Live Register

The operational front door of the translational pipeline: structured intake profile (business objective, autonomy degree, data sensitivity, deployment context, target users) → automated tier proposal (detectors + evaluator pipeline) → risk-proportionate approval workflow → register entry with AI-BOM stub. Prevents both over-engineering (blanket high-tier controls breed Shadow AI) and under-engineering (unassessed high-risk deployment). Every governance framework assumes it; almost no failed audit had one.

• article obligationArt. 17 — Quality Management System → AI Intake Portal & Use-Case TriageArt. 17 — Quality Management SystemQMS intake & change-approval workflow

Continuous, automated discovery of models, agents and LLM API calls across source repositories, cloud accounts, the CMDB and the employee SaaS footprint, reconciled into the AI register rather than collected by survey. It is the population-finding element: every inventory, assessment and vendor-governance duty is scoped against a set of systems, and a self-declared set is systematically short. Testable: run discovery against an estate containing one deliberately unregistered LLM integration and confirm it appears in the register within one sweep, attributed to an owner.

• control objectiveArt. 26 — Deployer Obligations → CO: Embedded-AI Vendor Governance → Shadow-AI Discovery & Asset Inventory ScannerArt. 26 — Deployer ObligationsCO: Embedded-AI Vendor GovernanceEmbedded-AI vendor governance presupposes a discovered population, not a declared one.

Jurisdictional isolation as an architectural boundary: compute, storage, backups, keys, operational metadata and control-plane administration held inside the target jurisdiction and operated by personnel subject to its law. The honest caveat is the control plane and metadata — a workload can sit in-region while telemetry, support access or identity services do not. Attestation surface: BSI C5, BSI C3A, ANSSI SecNumCloud.

• evidence artefactArt. 15 — Accuracy, Robustness, Cybersecurity → C5 / AIC4 Cloud Attestation → produced by Sovereign Deployment BoundaryArt. 15 — Accuracy, Robustness, CybersecurityC5 / AIC4 Cloud Attestation

Takes integrity proofs out of the operator's trust domain: periodic anchoring of log hash-chain heads via qualified electronic timestamps or a (qualified) electronic ledger per eIDAS 2, with signing keys held outside the operating team (key ceremony, HSM, separation of duties). Answers the insider test — a party who controls the vault cannot rewrite history without the anchor exposing it. Cost profile: anchoring is periodic and cheap; it upgrades every downstream log-based artifact at once.

• article obligationArt. 12 — Record-Keeping / Logging → External Trust Anchor (Qualified Timestamp / Ledger)Art. 12 — Record-Keeping / Loggingintegrity rung of the logging obligation
• control objectiveArt. 12 — Record-Keeping / Logging → CO: Log Integrity & Non-Repudiation → External Trust Anchor (Qualified Timestamp / Ledger)Art. 12 — Record-Keeping / LoggingCO: Log Integrity & Non-Repudiationthe rung the vault alone cannot reach

The marketplace/discovery API through which external agents find, authenticate against and transact with your agents: published capability descriptors, counterparty authentication, per-counterparty rate and value limits, and a resolvable record of which external principal initiated which transaction. Without it, business-to-agent traffic is anonymous inbound automation.

• article obligationArt. 25 — Value Chain / Role Flip → Agent Discovery & Registry EndpointArt. 25 — Value Chain / Role Flipattribution across the value chain when an external agent transacts