Customer developer docs

NIST AI Risk Management Framework (AI RMF 1.0)

NIST AI RMF 1.0 — GOVERN / MAP / MEASURE / MANAGE functions via policy snapshots, model cards, lineage, bias audits, and incidents.

NIST AI Risk Management Framework (AI RMF 1.0)

The NIST AI RMF is the canonical U.S. voluntary framework for managing AI risks. This mapping covers the four core functions — GOVERN, MAP, MEASURE, and MANAGE — with their key sub-categories. Ledgix evidences each function via policy snapshots, training-data lineage, model cards, dataset sheets, bias audits, and the operational ledger.

Status: Full — every control resolves to an artifact Ledgix produces today following the Phase 2 (model cards, dataset sheets), Phase 3 (bias audits), and Phase 9 (training data lineage) shipping.

Scope

AI RMF 1.0 is voluntary but widely adopted by U.S. federal agencies, enterprise AI governance programs, and is explicitly referenced by several U.S. state AI laws. Coverage here spans GOVERN (legal/regulatory policy mapping, risk management process), MAP (context and intended use, data provenance, stakeholder identification), MEASURE (trustworthy characteristics monitoring, fairness and bias detection), and MANAGE (risk response and recovery, post-deployment monitoring and change tracking).

Controls covered

FieldTypeRequiredDescription
NIST-AI-RMF-GOVERN-1.1policy_snapshots / events_jsonlGOVERN — Legal and Regulatory Policy MappingVersioned policy snapshots; every decision references the policy version that governed it.
NIST-AI-RMF-GOVERN-1.4policy_snapshots / framework_mappingGOVERN — Risk Management Process DocumentationPolicy lifecycle plus the framework mapping document itself.
NIST-AI-RMF-MAP-2.1events_jsonl / policy_snapshots / model_cards / dataset_sheetsMAP — AI System Context and Intended UseTool/action inventory, policy text defining allowed intents, model cards, and dataset sheets.
NIST-AI-RMF-MAP-3.1training_data_lineage / model_cards / dataset_sheetsMAP — Data Provenance and Lifecycle ContextSigned lineage records per model_ref plus internal model cards and dataset sheets.
NIST-AI-RMF-MAP-3.4events_jsonlMAP — Identification of AI System Stakeholdersagent_id and human_principal identify AI and human stakeholders per action.
NIST-AI-RMF-MEASURE-2.7events_jsonlMEASURE — Trustworthy Characteristics MonitoringPer-decision confidence, reasoning, citations, and evidence chunks operationalise explainability and reliability.
NIST-AI-RMF-MEASURE-2.11bias_audits / events_jsonlMEASURE — Fairness and Harmful Bias DetectionSigned per-window bias audit reports with four-fifths, chi-square, and p-value per protected-attribute group.
NIST-AI-RMF-MANAGE-2.3events_jsonl / incidentsMANAGE — Risk Response and RecoveryDenials, HITL overrides, and signed incident records document risk-response actions.
NIST-AI-RMF-MANAGE-4.1checkpoint_chain / key_history / incidentsMANAGE — Post-Deployment Monitoring and Change TrackingMerkle checkpoint chain, key rotation history, and post-deployment incident records.

Evidence types referenced

  • policy_snapshots — versioned governance and risk-treatment policy text.
  • events_jsonl — per-decision context, confidence, reasoning, and stakeholder identification.
  • model_cards — signed model cards enumerating intended use, performance, limitations.
  • dataset_sheets — dataset composition and known gaps.
  • training_data_lineage — signed lineage per model reference.
  • bias_audits — per-window signed bias audit reports.
  • incidents — signed incidents capturing detection, root-cause, and corrective action.
  • checkpoint_chain — Merkle checkpoint chain proving continuous monitoring.
  • key_history — key rotation history for long-horizon auditability.
  • framework_mapping — control-to-evidence correspondence document.

Known gaps (if any)

None — every control resolves to an artifact Ledgix produces today. Demographic stratification in MEASURE-2.11 requires tenants to tag events with subject_context.

Audit pack workflow

Export an evidence ZIP for this framework from the admin console's Evidence Exports panel by selecting NIST AI Risk Management Framework (AI RMF 1.0) and a time window. Each control's evidence_locators[] in the included framework_mapping.json points to the corresponding file in the ZIP.

References