UNESCO Recommendation on the Ethics of Artificial Intelligence
The UNESCO AI Ethics Recommendation (adopted 2021 by all 193 Member States) articulates ten values and principles for AI systems across their lifecycle — from human rights to sustainability and literacy. Ledgix evidences each value with the ledger, policy snapshots, cryptographic custody, and the framework mapping document.
Status: Full — every value resolves to an artifact Ledgix produces today. Fairness and privacy values are supplemented by the Phase 3 bias audits and Phase 5 DSR workflow for tenants that enable them.
Scope
The UNESCO Recommendation applies voluntarily to all 193 UNESCO Member States and to AI actors operating within them. Coverage spans ten values: human rights and dignity, proportionality and do-no-harm, safety and security, fairness and non-discrimination, sustainability, privacy and data protection, human oversight, transparency and explainability, responsibility and accountability, and awareness and literacy.
Controls covered
| Field | Type | Required | Description |
|---|---|---|---|
| UNESCO-AI-V1 | policy_snapshots / events_jsonl | Respect, protection and promotion of human rights and human dignity | Policies capturing rights-respecting rules; denied actions represent rights-protective enforcement. |
| UNESCO-AI-V2 | events_jsonl | Proportionality and do no harm | action_category and reason demonstrate proportionality assessment per decision. |
| UNESCO-AI-V3 | checkpoint_chain / key_history | Safety and security | Tamper-evident chain and cryptographic key custody evidence. |
| UNESCO-AI-V4 | events_jsonl | Fairness and non-discrimination | Aggregate fairness baselines by agent/action; demographic bias audits available for opt-in tenants. |
| UNESCO-AI-V5 | events_jsonl | Sustainability | Operational volume and time-series for downstream sustainability reporting. |
| UNESCO-AI-V6 | policy_snapshots / events_jsonl | Right to privacy and data protection | Policies governing data access; data-access decisions recorded with rationale. Subject-level privacy ops via Phase 5 DSRs. |
| UNESCO-AI-V7 | events_jsonl | Human oversight and determination | HITL-reviewed decisions and delegated-human-principal evidence. |
| UNESCO-AI-V8 | events_jsonl | Transparency and explainability | Per-decision reason, citations, evidence chunks. |
| UNESCO-AI-V9 | events_jsonl | Responsibility and accountability | Cryptographic attribution per decision (agent_id, receipt_signature, receipt_key_id). |
| UNESCO-AI-V10 | framework_mapping | Awareness and literacy | Exportable, human-readable mapping document communicates AI control coverage to non-technical stakeholders. |
Evidence types referenced
- policy_snapshots — rights-respecting and privacy-governing policy text.
- events_jsonl — per-decision reasoning, accountability metadata, and operational telemetry.
- checkpoint_chain — tamper-evident chain supporting safety-relevant forensic analysis.
- key_history — cryptographic key custody evidence.
- framework_mapping — human-readable map for public awareness and literacy.
Known gaps (if any)
None — every value resolves to an artifact Ledgix produces today. Demographic stratification for Value 4 (fairness) requires tenants to tag events with subject_context; subject-level Value 6 (privacy) operations use the Phase 5 DSR workflow.
Audit pack workflow
Export an evidence ZIP for this framework from the admin console's Evidence Exports panel by selecting UNESCO Recommendation on the Ethics of Artificial Intelligence and a time window. Each control's evidence_locators[] in the included framework_mapping.json points to the corresponding file in the ZIP.
References
- Framework mapping JSON:
vault/internal/compliance/frameworks/unesco_ai_ethics.json - Canonical source: UNESCO Recommendation on the Ethics of Artificial Intelligence — unesco.org