Security
US Government Pressures OpenAI and Anthropic to Restrict AI Models from the UK
The US government has intensified its pressure on OpenAI and Anthropic to restrict access to their most advanced AI models from the United Kingdom. This move signifies a notable shift in how the US approaches cutting-edge artificial intelligence, treating it more like a national security asset than a conventional software product.
In June 2026, the US Commerce Department issued an export-control directive targeting Anthropic’s Fable 5 and Mythos 5 models, citing cybersecurity risks related to potential 'jailbreaks' that could undermine safety protocols. This directive led to a complete suspension of foreign access to these models, affecting all users outside the US, including Anthropic’s own overseas employees.
OpenAI followed suit, restricting access to its GPT-5.6 Sol model exclusively to customers approved by the US government. Although OpenAI has indicated that it does not intend for this gatekeeping to be a permanent policy, compliance with the directive has left little room for alternative actions.
The situation escalated further by September 2026, when Anthropic denied pre-release testing access of its Claude Mythos 5.1 model to the UK's AI Security Institute (AISI), which had previously enjoyed such access as part of established collaborative safety-testing arrangements. Instead, access was offered only to vetted US organizations.
The UK government responded with a lobbying campaign seeking an exemption from these restrictions, arguing that its strong AI safety infrastructure and the Five Eyes intelligence-sharing relationship warranted special treatment. However, US officials reportedly responded that there was 'zero chance' of an exemption being granted, indicating a firm stance on the matter.
This development raises concerns about the implications for the UK’s access to advanced AI technologies and the competitive positioning of companies like Anthropic, especially as the market anticipates significant advancements in AI capabilities.
FAQ
Why is the US government pressuring OpenAI and Anthropic regarding their AI models?
The US government is treating advanced AI models as national security assets and has issued export-control directives due to cybersecurity risks associated with potential 'jailbreaks' that could undermine safety protocols.
What specific actions have OpenAI and Anthropic taken in response to the US government's directives?
Anthropic suspended foreign access to its Fable 5 and Mythos 5 models, while OpenAI restricted access to its GPT-5.6 Sol model exclusively to customers approved by the US government.
How has the UK government reacted to these restrictions?
The UK government has launched a lobbying campaign seeking an exemption from the restrictions, arguing that its AI safety infrastructure and intelligence-sharing relationships justify special treatment.
What was the outcome of Anthropic's decision regarding access to its Claude Mythos 5.1 model?
Anthropic denied pre-release testing access to the UK's AI Security Institute, which had previously enjoyed such access, and instead offered it only to vetted US organizations.
What are the potential implications of these restrictions for the UK?
The restrictions could limit the UK's access to advanced AI technologies, impacting the competitive positioning of companies like Anthropic and potentially hindering advancements in AI capabilities within the UK.