Graduated Boundary Enforcement
Objective
Ensure autonomy boundary enforcement is graduated (warning/restricted/prohibited zones) rather than a single binary safe/unsafe response.
Control / requirement
At least three zones defined (warning, restricted, prohibited), each with a distinct response — operator alert, autonomy reduction, hard stop — rather than one shared response.
Applicability
Systems where a single hard-stop response to any boundary approach would itself introduce risk (e.g., a moving vehicle, a drone in flight).
Expected evidence
Zone definition documentation; response-by-zone test records. [T3]
Assurance expectation
Test evidence demonstrating each zone triggers its own distinct, correct response, not a shared fallback.
Dependencies
This control terminates in a Domain 5 (Safe-State) response at its final zone — flagged honestly as the borderline case in the Domain 4 stress-test verdict below. Also adjacent to PAI-SF-ENV-002 (Domain 12) — see Open Items.
Exclusions
Not applicable to systems where any boundary approach is itself unacceptable risk and only a single hard-stop response is appropriate.
Maturity / conformance relevance
Expected at Operational and High-Assurance levels for systems where graduated response is safer than immediate stop.
Ecosystem relationship
Provides graduated-boundary-violation incident categories for UAIF™; the final-zone response is the AI- IRF™/Domain 5 handoff point.