Anthropic suspends its most powerful AI models following pressure from the US government

Anthropic has suspended access to its two most powerful artificial intelligence models, Fable 5 and Mythos 5, to comply with a US government directive citing ‘national security’, just three days after their commercial launch. Washington has ordered, under export control regulations, that access to these models be suspended for ‘any foreign national, inside or outside the US’, including Anthropic’s ‘foreign employees’, according to the AI startup. Believing it impossible to filter its users by nationality, Anthropic was forced to announce the suspension.

Anthropic, which a few days ago filed a confidential prospectus with the SEC ahead of its listing on Wall Street, received the order from the US government on Friday afternoon at 17:21 Washington time (23:21 in Italia). ‘The direct consequence of the order is that we must immediately disable Fable 5 and Mythos 5 for all our customers, in order to ensure compliance,’ the company reported, specifying that all other models will not be affected. Fable 5 and Mythos 5 are the two models that Anthropic has presented as cutting-edge in various industry benchmarks. Fable 5, in particular, marked the first time the start-up has made such an advanced offering public, thanks to new security measures that block responses in specific high-risk areas. The models were based on the release of Claude Mythos Preview, which in April had impressed Wall Street and government officials with its advanced cybersecurity capabilities. At the time, the company stated that it did not intend to make the model available to the general public, limiting its launch to an initial select group of companies as part of a cybersecurity initiative called Project Glasswing.

In its statement on Friday, Anthropic, whilst apologising to customers for the disruption, noted that, ‘as we have stated publicly, we believe the government should have the power to block unsafe applications, as part of a regulatory process that is transparent, fair, clear and based on concrete technical data’, while noting, however, that “this action does not respect those principles”. The announcement marks the latest clash between Anthropic and the US government. The Pentagon classified the start-up this year as a risk to the supply chain, believing the company poses a threat to US national security: a label, reserved for foreign adversaries, which requires defence contractors to certify that they will not use Anthropic’s Claude models in activities carried out for the armed forces. Anthropic has sued the Trump administration to overturn the measure in a legal dispute that is still ongoing.