Independent Physical Override
Objective
Ensure a physically accessible override exists, independent of the AI compute/software stack.
Control / requirement
Hardwired where kinetic risk is significant; logged, non- repudiable activation; authenticated reset process.
Applicability
All systems where humans are present in the operational environment or remote intervention is feasible.
Expected evidence
Architecture independence analysis; wiring diagrams; override activation/reset logs; test records under simulated AI compute failure. [T2/T3]
Assurance expectation
Physical demonstration of override function under failure conditions, not only architecture review.
Dependencies
PAI-SF-SAF-001 (override triggers safe state).
Exclusions
For systems with no human access, override may be remote with equivalent independence requirements.
Maturity / conformance relevance
Expected at all conformance levels for systems with human proximity.
Ecosystem relationship
Extends GAISSF™ human-in-the-loop controls; provides override-failure incident categories for UAIF™; defines override as primary AI-IRF™ response mechanism.