The US government has mandated that AI developer Anthropic restrict access to its latest models, Fable 5 and Mythos 5, for non-American nationals, citing national security concerns. This directive has prompted the multi-billion-dollar company, known for its Claude AI, to suspend these advanced models for all users globally, rather than implement a partial ban. The decision has sent ripples through the international technology community, raising significant questions about access to cutting-edge AI and the implications for national sovereignty, particularly in the UK.
Anthropic stated that the US government's order was based on concerns over a 'method of bypassing' the AI model, suggesting a potential vulnerability that users might exploit for harmful objectives. While the company acknowledged 'minor vulnerabilities' in its Fable model, it also asserted that safeguards had been implemented following collaborative cybersecurity efforts with both the US government and the UK's AI Security Institute. Anthropic maintained that no testers had yet discovered a 'universal jailbreak' capable of broadly circumventing the model's safeguards for a wide range of cyber capabilities.
Despite disagreeing with the breadth of the ban, Anthropic confirmed its compliance with the legal order. The firm also suggested that if these stringent standards were applied consistently, more AI models could face similar restrictions in the future. The Fable model, in particular, had been highly anticipated by users due to its advanced capabilities, including the creation of games, coding software, and control systems, making its suspension a notable setback for the AI development community.
The US administration's move has elicited strong reactions from politicians and tech analysts in the UK. Kanishka Narayan, the UK's AI Minister, posted on X, emphasising the critical importance of access to AI capabilities in the ongoing debate around national security and technological sovereignty. He also defended the UK government's investment in the UK Sovereign AI unit, an initiative aimed at funding homegrown AI founders to bolster domestic capabilities.
Former Security Minister Tom Tugendhat echoed these sentiments, describing the ban as an 'inevitable result of technology shaping warfare', where sovereignty increasingly hinges on 'code rather than cannons'. He criticised Britain's response, suggesting that high energy costs and an emphasis on safety over opportunity have led the UK to 'build the brake', potentially cutting itself off from future advancements. Raoul Ruparel, a former UK government adviser, termed the events a 'wake-up call' for nations beyond the US and China, warning of potential emerging security and economic disparities. Reform UK's Zia Yusuf also weighed in, claiming he had warned about such access bans for months and criticised the government for having 'virtually zero sovereign AI capability'.