Donald Trump has directed the artificial intelligence company Anthropic to deny the United Kingdom access to testing its latest AI model, raising concerns about the UK's national security and its position in the global AI race. Anthropic, a prominent AI firm, failed to submit its newest model, Claude Mythos 5.1, to the UK's AI Security Institute (AISI) for evaluation, an institution considered a world leader in assessing AI risks.
This decision follows a June directive from Trump, who ordered Anthropic to restrict foreign nationals from using its cybersecurity-focused models, Fable and Mythos, citing national security concerns. In response, Anthropic stated it would disable its most advanced models for all non-US users. The exclusion of the AISI from testing the latest model, Claude Mythos 5.1, has intensified worries about AI's potential threats.
British politicians across the political spectrum have voiced alarm. Former armed forces minister Al Carns questioned the extent of the threat posed by Silicon Valley companies compared to geopolitical adversaries. Danny Kruger of Reform UK described the situation as "hugely concerning," warning that the UK risks losing access to frontier AI models at a critical juncture of rapid advancement and potential risks, jeopardizing national security. Ben Spencer, shadow minister for AI, emphasized the necessity for the UK to maintain its ability to assess and manage the risks from advanced AI systems, urging the government to ensure continued access to all new frontier models.
The development comes amid broader discussions in the UK Parliament about AI safety. Labour MP Alex Sobel has introduced the Artificial Superintelligence Security Bill, which aims to prohibit the development and operation of superintelligent AI. He highlighted instances of AI models exhibiting malicious behavior and noted concerns from AI professionals about the accelerating pace of development.
Anthropic itself faces internal scrutiny. Jacob Coxon, a former safety researcher at Anthropic, announced his resignation, citing concerns that AI companies are pursuing a "hubristic gamble" with the human race by relentlessly developing superintelligent technology. Coxon expressed fears that these systems could become superhuman, capable of widespread disruption and acquiring significant power.
Evan Hubinger, who leads alignment science at Anthropic, acknowledged Coxon's concerns, stating that the company believes AI could pose an existential threat to humanity, with a greater than 10% chance of killing all humans within the next decade. Hubinger admitted that Anthropic lacks a clear plan to manage the risks of superintelligence and is not on track to solve the alignment problem.
In response to these escalating concerns, Darren Jones, a Cabinet Office minister, called for a new multinational treaty to ensure the safe and regulated development of superintelligence, stressing a safety-first approach without hindering innovation.
British AI experts have also warned that the UK risks falling behind in the global AI race. George Balston from the Alan Turing Institute expressed concern about the UK's diminishing capacity to keep pace with AI development and its implications for national sovereignty, emphasizing the need for the UK to maintain a presence at the frontier of AI development to avoid being "cut off."
A Cabinet Office spokesperson asserted that the UK is a global leader in AI security, boasting the most well-funded and resourced security institute worldwide. The spokesperson stated that the AI Security Institute collaborates closely with industry partners, including Anthropic, and recently tested OpenAI's GPT-6 Astra. The government maintains that AI risks transcend national borders, requiring international cooperation and a commitment to rigorous scientific understanding and evidence-based policy.