AI governance

DenialOS contains an explicit AI control plane.

Policy levels

Autonomy levels in the current implementation include:

L0, L1, L2, and higher governed levels.

The decision engine can produce:

• OBSERVE
• RECOMMEND
• PREPARE
• REQUIRE_HUMAN
• DENY
• KILL_SWITCHED.

Higher autonomy is not equivalent to unrestricted external execution.

Evaluation gates

The platform can record evaluation results by dimension and assess measurable safety gates. A capability may remain in a human-authority-only posture even when evaluation gates are met.

Kill switches

There are deployment-level AI kill switches and capability-level controls. These are intended to fail closed rather than silently grant authority.