Section 83 of 102
82. Independence of Safety-Critical Controls
Stable section ID: S05-CON-010-SECTION-83 · 15 content blocks
Safety-critical controls shall retain appropriate functional independence from general-purpose AI agents.
Independent systems may include:
- Fire detection.
- Fire suppression.
- Emergency communication.
- Electrical protection.
- Pressure and temperature protection.
- Emergency shutdown.
- Egress systems.
- Structural warning systems.
- Gas detection.
- Critical access and responder controls.
The AI agent may receive information from or provide bounded assistance to such systems, but it shall not become their sole means of detection, decision, or actuation unless specifically qualified under the applicable safety framework.
Failure, compromise, disconnection, or suspension of the agent shall not prevent a safety-critical system from performing its required function.
Information from the agent shall not silently change certified thresholds, logic, or protection settings.