Why 'Turning Off' AI Might Be Harder Than It Sounds

New proposed legislation could require organisations using AI agents to have the ability to throttle, suspend, or fully shut them down if something goes wrong. The idea of an 'AI kill switch' sounds simple in theory, but experts note that deciding exactly when and how to pull that switch is far more complicated in practice.

For businesses increasingly relying on AI tools for customer service, automation, or decision-making, this raises important questions: what counts as a failure serious enough to warrant shutdown, who has the authority to make that call, and how quickly can it be done without disrupting operations? These are not just technical questions but governance ones, and many organisations have not yet thought through the answers.

As AI regulation develops, small and medium businesses should not wait for laws to force their hand. Understanding what AI systems are being used internally, who controls them, and how they can be paused or disabled if they misbehave is a basic risk management step that applies regardless of company size.