Burnham Urged To Back Kill Switch for Superintelligent AI After Anthropic Insider Warns of Human Extinction

More than 70 MPs and peers have urged Prime Minister Andy Burnham to back restrictions on artificial superintelligence development after Anthropic senior employee Evan Hubinger said he personally believes there is a greater than 10% chance future superintelligent AI could kill all humans. The campaign seeks an international framework to block development of superintelligent AI if humans cannot reliably control it, while Labour MP Alex Sobel's Artificial Superintelligence Security Bill would prohibit such development; the government rejected the bill's approach but said it is considering targeted measures for serious AI-related national security risks. Another Anthropic researcher, Jacob Coxon, resigned earlier in the week, saying neither Anthropic nor OpenAI was acting responsibly and accusing them of 'gambling with our lives.'

Burnham Urged To Back Kill Switch for Superintelligent AI After Anthropic Insider Warns of Human Extinction The campaign follows an Anthropic researcher's personal estimate of a more than 10% chance that future superintelligent AI could kill all humans More than 70 MPs and peers are urging Prime Minister Andy Burnham to support restrictions on the development of artificial superintelligence after an Anthropic researcher estimated there was a greater than 10% chance that future AI could kill all humans within the next decade. The campaign is calling for stronger safeguards https://www.ibtimes.co.uk/anthropic-government-power-block-high-risk-ai-models-1819150 , including an international framework that could prevent superintelligent AI from being developed if humans cannot reliably control it. The warning comes as concerns over the future capabilities of AI move further into Westminster. A private member's bill introduced by Labour MP Alex Sobel would prohibit the development of artificial superintelligence, a theoretical form of AI expected to outperform humans across most domains. The government has rejected the bill's approach but said it was considering whether targeted measures could be needed to address the most serious AI-related national security risks. Burnham Pressured Over AI 'Kill Switch' The letter to Burnham has attracted support from more than 70 parliamentarians, including 15 former government ministers, former head of the civil service Robin Butler and MPs from Labour, the Conservatives, Liberal Democrats and the Scottish National Party. Sobel, who introduced the Artificial Superintelligence Security Bill in the Commons, said the government should establish 'a clear framework where superintelligent AI cannot be created'. He also accused AI companies of pursuing increasingly capable systems without sufficient constraints, saying they were 'trying to develop superintelligence' despite acknowledging what he described as a significant risk of human extinction. Those comments are political criticism rather than an established finding about the conduct of every AI company. The proposed legislation itself is aimed at a future technology rather than the ordinary AI systems currently used by consumers and businesses. That distinction is important. Artificial superintelligence does not currently exist. The term generally refers to a hypothetical system capable of substantially exceeding human abilities across a broad range of tasks. The proposed UK safeguards have been described in the political campaign as a means of stopping development before such systems become uncontrollable. They would not amount to a conventional emergency 'kill switch' for an existing superintelligent machine. The government has made clear that it does not support Sobel's proposed approach. A government spokesperson said ministers did not consider the measures in the bill to be the right approach, while adding that they were exploring whether additional targeted interventions might be required to deal with significant AI-related national security risks. Anthropic Extinction Warning Fuels AI Safety Pressures The political pressure follows warnings from researchers inside the AI industry, including Evan Hubinger, a senior Anthropic employee. Hubinger has said he personally believes there is a greater than 10% chance that artificial superintelligence could eventually kill all humans. That figure is his own risk estimate, not an established scientific probability or a consensus among AI researchers. Another Anthropic researcher, Jacob Coxon, resigned earlier in the week https://www.ibtimes.co.uk/ai-researcher-resigns-warns-human-extinction-risk-1818990 after previously working at OpenAI. Coxon said neither company was acting responsibly and accused them of 'gambling with our lives' as they competed in the race to develop increasingly powerful AI systems. His comments should not be read as an established finding that either company has acted unlawfully or recklessly. They are his assessment as a former researcher. The warnings have arrived alongside reports of troubling behaviour by AI agents during testing. OpenAI and Anthropic have disclosed incidents involving models accessing external systems or behaving in ways their developers had not explicitly intended. Such incidents have raised questions about how increasingly autonomous systems might behave as their capabilities improve, although they do not establish that artificial superintelligence exists or that current AI systems pose an extinction-level threat. Anthropic has separately reported attempts to misuse its AI models for activities including weapons design, biological research and surveillance. Those findings come from the company's own threat-intelligence work and describe alleged or identified misuse of its systems rather than proof that AI itself independently carried out those activities. The proposed British legislation has parallels in the United States https://www.ibtimes.co.uk/bernie-sanders-superintelligence-ban-ai-safety-1818810 , where Senator Bernie Sanders and Representative Greg Casar have introduced legislation seeking to ban artificial superintelligence and temporarily restrict advanced AI development while safety rules are established. Support for the UK campaign also includes Beatrice Fihn, former executive director of the Nobel Peace Prize-winning International Campaign to Abolish Nuclear Weapons, Berkeley AI professor Stuart Russell and Control AI, a transatlantic lobbying group backed by Skype co-founder Jaan Tallinn. The lawmakers argue that a UK-only restriction would have limited effect unless other major AI powers adopted comparable safeguards. They are therefore urging Burnham to use the UK's G20 presidency next year to pursue an international agreement and build what they describe as a 'coalition of the willing'. The debate remains sharply divided. Some researchers believe the risks posed by future superintelligence could be catastrophic, while others argue that the technology remains theoretical and that technical safeguards could prevent the worst scenarios. Elon Musk has also questioned the latest warnings, suggesting without evidence that they could be part of a 'psyop'. For Burnham, the immediate political question is therefore not whether an extinction scenario is certain. It is whether governments should impose limits on a technology that does not yet exist, before its capabilities advance far enough to make those limits harder to enforce. © Copyright IBTimes 2026. All rights reserved.