Kill Switch
A mechanism to immediately stop or disable an AI system when it produces harmful, unsafe, or unauthorized outputs. Kill switches range from simple on/off controls to sophisticated graduated responses that can throttle, redirect, or degrade AI functionality without full shutdown.
Why It Matters
When an AI system goes sideways in production, every minute of uncontrolled operation compounds the damage. A kill switch is the emergency brake that prevents a bad model from becoming a catastrophic incident.
Example
A trading firm's AI system has a kill switch that automatically halts all algorithmic trades if the model's predictions deviate from expected ranges by more than three standard deviations, preventing a flash crash scenario.
Think of it like...
A kill switch is the big red emergency stop button on factory machinery — you hope you never need it, but when something goes wrong, it's the most important control in the building.
Related Terms
AI Incident
An event where an AI system causes or nearly causes harm, produces unintended outputs, or fails to perform as expected in ways that affect individuals, organizations, or the public. AI incidents require documented response, root cause analysis, and may trigger regulatory reporting obligations.
Human-in-Command (HIC)
A governance principle where humans retain ultimate authority and control over AI systems, including the ability to decide the scope of AI autonomy, override any AI decision, modify the system's behavior, and shut it down entirely. HIC is the overarching principle that encompasses both HITL and HOTL as implementation patterns.