Anthropic Withholds New Claude Model From UK Safety Testing

Anthropic did not give the UK's AI Security Institute pre-release access to its newest frontier model, Claude Mythos 5.1, breaking with a practice that had seen the agency test every previous Anthropic system before launch, according to a report citing people familiar with the matter. The model was released on September 1 and initially limited to vetted US organizations, while comparable American bodies were granted access. Anthropic has not publicly explained the decision. The Financial Times first reported the exclusion, and UK officials have raised concerns that it may reflect a wider protectionist shift among US AI companies aligned with the Trump administration. There is no public evidence that the US government ordered Anthropic to withhold the model. The White House has said the administration does not formally sign off on products released by private-sector firms and that companies are responsible for the timing and scope of voluntary testing. The Cabinet Office did not confirm that Mythos 5.1 had been withheld.