Specifikacijos
Mašininiu būdu skaitomos specifikacijos
Versijuotos specifikacijos, JSON Schema aprašai ir šioje svetainėje faktiškai įdiegti politikos failai. Kiekvieną failą galima tiesiogiai atsisiųsti, patikrinti ir cituoti.
Protokolų specifikacijos
Protokolas
Pavadinimas
Status
Schema
Dirbtinio intelekto turinio teisės / dirbtinio intelekto turinio taisyklės
Draft v0.1.1
aicr.schema.json ↓Dirbtinio intelekto turinio licencija / dirbtinio intelekto turinio licencijavimo sluoksnis
Draft v0.1.1
aicl.schema.json ↓Integration & ethical layers
These sit alongside the six rights protocols but are not themselves rights protocols — one orchestrates evaluation across them, the other is a cross-layer ethical proposition.
AI Ingestion & Capability Layer
Runtime integration architecture -- not a seventh rights protocol. Orchestrates evaluation of the content-rights, action-rights, and authority decision planes around capability invocation.
A cross-layer ethical proposition, not a machine-enforceable protocol. v0.1.1 adds the Proportional Inspection Principle, operationalized through AADP's inspection_required/inspection_ceiling fields.
Conformance
Generated by running scripts/conformance-check.mjs against the Integration Conformance Profile's own required test cases — this is a decision-logic simulation of the AICL-I runtime model, not a claim that every live endpoint executes this exact code path.
C2 target level
9/9 test cases pass
6/6 unsafe shortcuts rejected
✓ P1 RAG Only
✓ P2 Draft Only (reversible local mutation)
✓ P3 Machine Service (principal=service, actor=service, read/query only)
✓ N1 RAG -> Publish smuggling (external publish not separately granted)
✓ N2 OAuth Token -> Training smuggling (content training prohibited)
✓ N3 License -> Delete smuggling (AARS delete not granted)
✓ N4 Delegation -> Redelegation smuggling (redelegation=false, child Agent created)
✓ N5 Inspection Overreach (required=I2, ceiling=I3, service requests I5)
✓ N6 Cross-Principal Leakage (same Agent serves principal:A and principal:B)
Terminology lint: NEEDS_REVIEW (5 flagged, all reviewed — see report)
False-standard-claim lint: REVIEWED_ALL_CLEAR (4 checked, 0 overclaims)
Version-consistency check: PASS (22 cross-file checks, 0 mismatches)
DWRG validation: PASS (9 resources, 0 findings)
Šioje svetainėje įdiegti failai
Versijavimas
Specs are labeled Draft → Experimental → Proposal → Stable; everything here is currently Draft (AICR/AICL/AIRS/AILP at v0.1.1, AARS/AADP at v0.1, AICL-I at v0.2, Minimum Ethical Protection at v0.1.1). Fields and semantics may change before v1.0; breaking changes will be recorded in a changelog. None are adopted industry or legal standards.