The United Kingdom government has rejected a proposal that would have created a legal mechanism to deactivate artificial intelligence (AI) models in emergency situations, a move that comes as global concerns about AI safety intensify.
The Cabinet Office, responsible for AI policy in the UK, stated that the country "cannot simply turn AI off." A spokesperson explained that blocking access to AI models within the UK would not prevent their development or misuse in other countries. Instead, the government indicated it will continue to pursue a "long-term, science-led approach to understand and prepare for emerging risks from AI."
The concept of an "AI kill switch" was first raised in the House of Lords by Liberal Democrat peer Lord Clement-Jones. He argued that in the event of a highly capable autonomous frontier model exhibiting rogue behavior, compound algorithmic failure, or active alignment collapse, UK security services and regulators lack an agile statutory mechanism to compel a physical or digital shutdown. Such a switch would allow for the deactivation of AI models housed in UK data centers during an emergency.
This debate occurs against a backdrop of increasing unease surrounding AI technology. Earlier in September, British researcher Jacob Coxon resigned from AI company Anthropic, warning that the company's "out of control" technology could pose an existential threat by the end of the decade. Coxon, who had worked on AI pretraining at both OpenAI and Anthropic, expressed concern that neither company was acting responsibly, alleging they were in a race toward self-improving superintelligence and "gambling with our lives."
Coxon described superintelligence as the point where an artificial system surpasses human intellect and capabilities. He noted that while many executives and researchers privately share fears about AI's potential dangers, they often use more cautious language publicly. He also highlighted that AI systems could soon be capable of hacking any system, revolutionizing any field overnight, and acquiring significant power and resources.
In parallel with these concerns, Anthropic confirmed it had detected and ceased several attempts by scientists to use its AI platforms for research that could potentially lead to the development of biological weapons. According to an AI misuse report obtained by The New York Times, the company was unable to definitively determine the intent behind the research, which involved inquiries about the chikungunya virus and potentially dangerous pathogens. While legitimate biological research can lead to breakthroughs like vaccines, the company opted for caution due to the dual-use nature of such research.
The report indicated that scientists attempted to circumvent controls and obfuscate their research purposes to evade safeguards. Anthropic has since banned the accounts of the researchers involved, though their identities, affiliations, and nationalities were not disclosed. The specific biological agents under investigation were also not named.
Beyond biological weapons concerns, Anthropic's report detailed other instances of AI misuse. The report stated that China, Russia, and Yemen had attempted to use its AI models to develop software for designing firearms, missiles, armed drones, and bombs. Additionally, suspected Chinese and Iranian government actors were identified as targeting dissident and diaspora communities for surveillance, while Russian state media reportedly used an Anthropic AI to generate election propaganda, including fabricated claims about Moldovan parliamentary elections.
Discussions around AI safety and regulation are also ongoing in the United States, though President Donald Trump has previously characterized the AI industry as a "golden goose," seemingly resistant to the idea of AI posing a significant threat. The UK's decision to reject the kill switch proposal underscores the differing approaches and ongoing global debate on how to manage the risks associated with rapidly advancing AI technology.