I6: Bounded Actuation and Escalation Discipline
invariantoriginal
ClaimThe stabilizer action set is a whitelist with known risk profiles; high-risk actions require opt-in and high confidence; escalation to human support is suggested when risk markers rise.
This claim fails if
If the system takes actions outside the whitelist or acts with high risk under low confidence, this invariant is violated.
The stabilizer action set is a whitelist with known risk profiles. High-risk actions require opt-in and high confidence. Escalation to human support is suggested when risk markers rise.
Test criteria
Action whitelist review; fail-safe mode defaults to silence or state legibility under high uncertainty.