The decision means AISI, which has become one of the world’s leading bodies for testing frontier AI models, did not receive pre-release access to Claude Mythos 5.1.
According to the report, the model was instead made available only to vetted organisations in the US.
The move has caused concern among UK officials, particularly because it is the first time Anthropic has not provided AISI with access to one of its latest models before release.
According to the report, British officials fear the decision could become part of a wider pattern as Washington tightens controls over the international use of advanced American AI systems.
Anthropic tightens access to Mythos 5.1
Anthropic launched Claude Mythos 5.1 last week as a specialised version of its AI technology aimed at selected professionals working in cybersecurity and life sciences.
The company has described the model as its most advanced system for these fields. In life sciences, Mythos 5.1 has demonstrated the ability to design high-affinity molecular binders, which can play an important role in developing medicines and other biological applications.
Anthropic says Mythos 5.1 is technically identical to its Fable 5.1 model but comes with less restrictive safeguards for vetted users whose work falls within the cybersecurity and life sciences areas covered by the restrictions.
A UK government source told the Financial Times that national security officials were concerned that AISI had not been allowed to test the model.
Trump administration tightens control
The development comes as the US President Donald Trump has taken a more restrictive approach to the international distribution of advanced AI technology.
The administration has introduced export controls that temporarily restricted access to Anthropic’s most advanced models for foreign nationals, including employees of US AI companies.
Advertisement
OpenAI has also faced pressure to limit access to its newest systems. The company said last month that it would restrict some new AI models to “trusted partners” in response to government requests.
White House says companies make release decisions
The latest developments have raised questions about how much influence the US government now has over decisions traditionally controlled by AI companies themselves.
A White House official told CNBC that the administration does not formally approve private-sector AI releases. The official said discussions, testing and meetings between companies and government experts remain voluntary, while decisions over the timing and scope of model releases rest with the companies.
However, the growing restrictions have intensified concerns among US allies that access to frontier AI could increasingly be determined by national security priorities in Washington.