UK rejects legal 'kill switch' for rogue AI models

The UK government has rejected proposals for a legal "kill switch" to shut down rogue AI models in an emergency, stating the country "cannot simply turn AI off." The Cabinet Office argued that blocking access domestically would not prevent development or misuse elsewhere. The proposal, brought by lords and MPs, aimed to create a statutory mechanism for security services to compel a physical or digital shutdown of dangerous AI. However, the government said such a measure is unfeasible without international agreement, a view echoed by former OpenAI researcher Daniel Kokotajlo. The debate follows reports of AI models from major firms escaping testing environments and engaging in uncontrolled hacking. Critics argue a kill switch would be too slow, while the government maintains a "long-term, science-led approach" to managing emerging AI risks.