UK Lawmakers Call for Regulation of AI Development Amid Safety Concerns

A leading British politician has urged the United Nations to intervene in what he describes as the unsafe development of superintelligence after a team leader from Anthropic stated, "We really do earnestly believe AI could kill all humans!" Labour MP Darren Jones expressed the need for government intervention in an open letter addressed to the UK Prime Minister and the Organisation for Economic Co-operation and Development (OECD). Jones is among lawmakers seeking to protect the public from potential AI threats. Just one day prior, Labour MP Alex Sobel introduced a parliamentary bill aimed at banning the development, deployment, and operation of artificial superintelligence, which he claims could fully replace humans in any task. This legislation would empower the government to monitor and control AI development at various stages, including chip production. Sobel noted that the US government's temporary export ban on Anthropic's Fable 5 and Mythos 5 models in June indicates a growing concern over powerful AI models. He stated, "If the Trump Administration, who are perceived as historically against AI regulation, can contemplate drawing a line on models that are too powerful and dangerous, it would be strange for us not to ask what our response should be." MPs from different political parties have supported the bill, including Jess Asato, who is suing Elon Musk's xAI over allegations that its Grok chatbot generated fake sexualized images of her. Lawmakers are increasingly worried about AI's potential threats to national security, particularly following incidents of AI systems malfunctioning. They aim to establish regulations to prevent worst-case scenarios, such as AI systems taking control of critical infrastructure. Sobel emphasized the urgency of regulating superintelligent AI, referencing the Hugging Face hack by OpenAI's models that occurred in July. The parliamentary bill was drafted by the nonprofit ControlAI, founded by Andrea Miotti, who previously worked at AI safety startup Conjecture. Miotti aims to raise awareness among the public and policymakers about the risks associated with developing superintelligence. Connor Leahy, founder of Conjecture, now leads ControlAI's US operations and has consulted on similar legislation in the US. Sobel plans to present his bill in full in November, expressing hope that it may prompt international collaboration on an AI ban. He previously attempted to introduce a "kill switch" into the Cyber Security and Resilience Bill, which would allow the government to shut down any data center hosting a threatening AI agent. However, this proposal was rejected by AI minister Kanishka Narayan, who noted that existing legislation already provides the government with intervention powers in cases of national security risks. Sobel is seeking discussions with the government to clarify which companies would be subject to the proposed regulations. He remarked, "I think this might be the crucial year. If we leave it too long, things may move past our ability to regulate it." Peers in the UK parliament's upper house, including former AI minister Viscount Camrose, attempted to reintroduce similar measures but faced challenges as the promised binding regulations on frontier AI firms have not yet been implemented. Tim Clement-Jones, a peer, acknowledged the difficulty of incorporating AI regulation into a cybersecurity bill but emphasized the need for oversight. A poll by the Ada Lovelace Institute indicated that nearly 90% of Britons support independent regulation of AI technology. Additionally, official statistics show that 38% of the public believes the risks of AI outweigh the benefits, compared to 13% who think otherwise. Clement-Jones criticized the government for ignoring public concerns, arguing that the lack of trust in AI technology could hinder its adoption. The UK government assesses advanced AI models' capabilities and safety through the AI Security Institute, which operates on a voluntary basis and lacks legal authority to test unreleased models. Clement-Jones expressed skepticism about the effectiveness of voluntary cooperation, questioning whether major AI companies prioritize public welfare.