Safety & Policy
Deterministic constraints on AI physical agency
Autonomy Levels (AL 0-4)
No Autonomy (Suspended)
Agent may process data but cannot propose or execute any actions. Read-only state.
Human Commanded
Agent may propose actions. ALL actions require explicit human approval prior to execution.
Bounded Autonomy (Standard)
Agent may execute pre-approved diagnostic sequences autonomously. State-changing actions (e.g., parameter shifts) require human approval.
Supervised Operation
Agent may execute state-changing actions within defined safety envelopes without per-action approval. Human may interrupt.
Full Autonomy
Agent has full authority over delegated subsystems. Governed only by deterministic hardware interlocks.
Global Constraints
Hardware hard-limit overriding all agents
For AL1 and AL2 state changes
Suspend agent on 3 consecutive policy blocks
Recent Blocks
Attempted UPDATE_WAVEFORM without AL3 authority.
Proposed parameter shift exceeded ±5MHz bounds constraint.