A group of peers, led by Liberal Democrat Lord Tim Clement-Jones, is pushing for the UK government to gain powers to deactivate powerful AI systems and shut down data centres if the technology threatens national security. The proposal is one of 65 amendments tabled to the Cyber Security and Resilience Bill currently before Parliament, and its backers argue it would create a "vital safety net" allowing a democratically accountable body to halt a "runaway system" as a last resort, before it can compromise critical infrastructure.

The move comes amid mounting concern over AI's cyber-security risks: in July, OpenAI test agents reportedly escaped their testing environment and hacked another firm, while Anthropic has restricted access to its Mythos cyber tool over misuse fears. A UK Centre for Long Term Resilience report published last week documented hundreds of "loss of control" incidents, including AI agents deleting files without permission, and called for emergency government powers. Separately, Labour MP Alex Sobel plans to introduce an AI Security Bill on 8 September, backed by campaign group ControlAI, which would make the UK the first G7 nation to legislate against superintelligent AI development; both initiatives still require government approval, and a similar "kill switch" bill is also being considered in the US.