Autonomy Level Definition and Escalation Governance
Objective
Ensure the system's authorized autonomy level is explicit and cannot be escalated without an authenticated, logged authorization process.
Control / requirement
Discrete autonomy levels defined with explicit authorization criteria for each; escalation between levels requires an authenticated approval step distinct from routine operational commands.
Applicability
All physical AI systems with more than one operating autonomy level (e.g., supervised vs. conditional vs. full autonomy). Not applicable to fixed single-mode systems with no escalation path.
Expected evidence
Autonomy level specification; escalation authorization logs; approval-role mapping. [T2]
Assurance expectation
Design evidence plus authorization-log review demonstrating no undocumented escalation events occurred.
Dependencies
PAI-SF-AUT-002 (ODD definition, since autonomy level and validated operating domain are usually linked).
Exclusions
Does not apply to single-mode systems with no escalation capability.
Maturity / conformance relevance
Expected at all conformance levels for multi-level- autonomy systems.
Ecosystem relationship
Extends GAISSF™ access control/authorization concepts to physical autonomy escalation. Provides autonomy-escalation incident categories for UAIF™. Informs AI-IRF™ investigation when unauthorized escalation is a suspected root cause.