Key facts
- The US is expected to reject the UK's request for access to Anthropic's advanced AI models, Fable 5 and Mythos 5.
- US officials cited security concerns that the models could be 'jailbroken' to unlock advanced cybersecurity capabilities.
- Anthropic disputes these concerns, stating they relate to a narrow vulnerability and could hinder frontier AI releases.
- The dispute highlights the UK's reliance on US technology providers and the need for sovereign AI capabilities.
- UK AI minister Kanishka Narayan and business secretary Peter Kyle emphasized the importance of domestic tech innovation.
- Cyber-security experts have warned that restricting access for researchers could make networks less secure.
The White House is poised to reject the UK's plea for an exemption to access Anthropic's advanced AI models, Fable 5 and Mythos 5, exposing Britain's dependence on a few US Big Tech providers. Downing Street has lobbied Washington for access after the Trump administration blocked foreign use of these powerful AI systems, citing concerns that the models could be 'jailbroken' to unlock advanced cybersecurity capabilities. Anthropic disputes these security concerns, arguing they are based on a narrow vulnerability and that such restrictions could halt the release of frontier AI models. Emergency talks between Anthropic and the US Commerce Department ended without lifting the restrictions. UK officials, including AI minister Kanishka Narayan and business secretary Peter Kyle, have stressed the need for the UK to develop its own 'sovereign AI capability' and accelerate domestic innovation. Keir Starmer is expected to discuss the matter with President Trump at the G7 summit, though country-specific exemptions are unlikely. The dispute has also drawn criticism from cyber-security experts who warn that restricting access for legitimate researchers could weaken Western networks.