US lets Anthropic release Mythos AI to ‘trusted’ firms

TL;DR:

• The US government has allowed Anthropic to redeploy its Claude Mythos 5 model to “trusted” US organisations, partially reversing a two-week-old suspension.
• More than 100 companies and institutions, many of them Fortune 500 firms, will regain access via Anthropic’s Project Glasswing.
• The vetting process has drawn criticism for its opacity, with even Sam Altman objecting to the government picking who gets access.

Anthropic said the US government has cleared it to release its powerful Claude Mythos 5 model to a set of “trusted” US organisations, softening an order issued two weeks ago that suspended access on national security grounds. Mythos is Anthropic’s strongest cybersecurity model, and the redeployment targets bodies that “operate and defend critical infrastructure”.

A vetted list, but little transparency

More than 100 organisations will regain access, a source told Reuters, many of them part of Anthropic’s Project Glasswing grouping of well-known tech firms and institutions. A letter from commerce secretary Howard Lutnick said export licences would no longer be required for Mythos 5 to reach approved companies and their non-US-citizen employees, though restrictions remain for firms off the list.

The selection process has been widely criticised. “No one knows how these companies are picked and why everyone else is excluded,” said John Coleman of the Foundation for Individual Rights and Expression, warning it concentrates too much power in government hands. OpenAI’s Sam Altman echoed the point even as his own firm accepted similar curbs on GPT-5.6.

The episode caps a fraught period for Anthropic, which had earlier been accused by the US of letting Mythos probe classified systems and faced a legal challenge over the export order. The Guardian noted the UK’s own AI security body has described Mythos as a “step up” over previous frontier models.

Looking forward

A wider release of Fable 5, the publicly available counterpart to Mythos, is said to be moving closer, though no timeline has been set. For UK organisations, the saga is a reminder that access to the most capable models is increasingly shaped by US clearance decisions made well outside their control — a sovereignty question that sits awkwardly alongside Britain’s ambitions to be a serious AI player.