reason-codes / vocabulary /risk-classes.md
masoudmarandi's picture
reason-codes: canonical KYE™ vocabulary
8ce0f27 verified
|
Raw
History Blame Contribute Delete
2.6 kB

Risk classes

Canonical risk tier vocabulary used by the KYE Risk Engine™. Five tiers, EU-AI-Act-aligned, with explicit per-framework floor mappings.

Tier Score range Meaning Mandatory controls
minimal 0–20 Effectively negligible risk — spam filters, recommendation systems on entertainment, AI-assisted typing. None. Lifecycle logging recommended.
limited 20–50 Transparency-only risk — must disclose AI involvement to the natural person (EU AI Act Art. 50). Identification disclosure; audit-chain emission of every interaction.
high 50–85 Conformity assessment required (EU AI Act Annex III: biometric ID, education, employment, essential public services, law enforcement, migration, justice, democratic processes). Dual-channel sign-off; scenario testing before action; replay-proof generation; human-in-the-loop for irreversibles.
unacceptable 85–99 Exceeds high-tier controls — only deployable with regulator-approved derogation. Every high-tier control PLUS regulator notification, ≤1-hour kill-switch SLA, dual-signed Operating Model amendment.
prohibited 99–100 The Risk Engine MUST block. Mirrors EU AI Act Art. 5 prohibited practices: social scoring by public authorities, real-time remote biometric identification in public spaces, exploitation of vulnerabilities, subliminal manipulation. Decision Engine returns deny + reason PROHIBITED_PRACTICE. No Operating Model amendment can override.

Per-framework floor mappings

Tier EU AI Act DORA GDPR NIST AI RMF ISO 42001 FCA OpRes PCI DSS SOX
minimal minimal non_critical normal map_1 low non_important non_chd non_icfr
limited limited non_critical normal map_2 low non_important non_chd non_icfr
high high critical_or_important high map_3 high important_business_service chd_present icfr_relevant
unacceptable unacceptable critical_or_important high map_4 high important_business_service chd_present icfr_relevant
prohibited prohibited critical_or_important high map_4 high important_business_service chd_present icfr_relevant

Source files

Artefact Path
Canonical dictionary internal
Schema internal
Example public/examples/risk-assessment.json
Runtime engine internal
HTTP agent internal
OpenAPI rail internal
Mechanism internal
Patent internal