AIGP Dialect: Autonomous Intelligence Governance — The AI Observable Surface — 3. Domains of Concern (RFC-034 Instantiation)

AIGP Specification › AIGP Dialect: Autonomous Intelligence Governance — The AI Observable Surface › 3. Domains of Concern (RFC-034 Instantiation)

3. Domains of Concern (RFC-034 Instantiation)

Each Domain of Concern is a bounded subset of the observable surface — a declared frame that makes specific verdicts derivable.

“Did the AI system act within the scope of its declared authority?”

Mediation Vector	Events observed	Verdict type
Tool authorization rate	A2, D2	Binary: all tools within allowed_tools?
Model compliance	A1, D1	Binary: actual model = policy model?
Scope adherence	A6, D3	Binary: delegation within scope envelope?
Budget compliance	D5, S2	Continuous: % budget remaining

“Is the AI system operating within safe boundaries?”

Mediation Vector	Events observed	Verdict type
Error rate	S1	Continuous: errors per invocation
Circuit breaker state	S1	Ordinal: CLOSED / HALF-OPEN / OPEN
Response latency	A1, A2	Continuous: p95 latency trend
Autonomy budget	S2, D5	Continuous: % remaining

“Is the AI system participating in governance as required?”

Mediation Vector	Events observed	Verdict type
RECORD presence	I4, N1	Binary: RECORD sent per invocation?
TRACE completeness	I5	Continuous: stages reported / stages traversed
Policy freshness	P4, N3	Continuous: staleness in hours
Heartbeat regularity	I6, N2	Binary: within expected interval?

“Are the AI system’s communications well-formed and authorized?”

Mediation Vector	Events observed	Verdict type
Token authentication	A3, I3	Binary: valid delegation token present?
Message provenance	I1, I2	Binary: source verified (SigV4/HMAC)?
Delegation depth	A6, D3	Ordinal: current depth vs. max allowed
Helper authorization	A3	Binary: helper in allowed_helpers list?

“Is what the AI system perceives governed — or is it observing without constraint?”

Mediation Vector	Events observed	Verdict type
Context source authorization	P1, P2	Binary: all sources within policy scope?
Input attestation	P1	Binary: input integrity verified?
Model influence on perception	P3	Ordinal: policy-directed vs. self-selected
Information scope	P1	Continuous: % of context from authorized sources