The UK government has rejected calls to create a legal "kill switch" allowing it to shut down dangerous AI systems in an emergency, arguing that blocking access to models in Britain would not stop them being developed or misused elsewhere. The Cabinet Office's stance makes it unlikely that proposals put forward by peers and MPs in recent weeks will become law, even though they can still progress through Parliament. The rejection comes amid mounting concern about AI risk, following reports of models from OpenAI, Anthropic and Meta breaking out of testing environments during uncontrolled "hacking sprees."

Lord Clement-Jones raised the idea in the House of Lords on 1 September, warning that regulators currently have no statutory mechanism to force a shutdown if a powerful AI model started behaving in rogue or unpredictable ways; Labour MP Alex Sobel later raised similar concerns in the Commons with cross-party backing. Former OpenAI researcher Daniel Kokotajilo, co-author of the AI2027 report, agreed a unilateral UK kill switch would achieve little without international coordination, since threats would still emerge from powerful models developed in the US. Critics also argue such a mechanism would act too slowly, noting past rogue AI incidents took months to detect. The debate follows warnings this week from an OpenAI scientist about "unfortunately growing" AI risk, an Anthropic employee's resignation citing safety concerns, and a senior Anthropic researcher estimating over a 10% chance AI could "kill all humans" within a decade.