A safety control that stops an AI system fast during an emergency.
It is the big red button by a factory machine. If things go sideways, you hit it. No polite meeting needed.
It stops high-risk AI systems before they cause more harm. People use it when normal safety rules are not enough.
AI Guardrails
The kill switch stops the system when normal safety rules fail.
Alignment
Alignment lowers the chance of trouble. The kill switch limits damage after trouble starts.
Agent Permissions
Permissions limit what an AI can do. This can reduce the need for an emergency stop.