Anthropic’s Mythos mess is only getting worse

The Trump administration approved Anthropic's Mythos 5 AI model for limited use on June 26th, after initially issuing an export control order on June 12th that demanded the company suspend access to the model due to security concerns.

The export control order, which covered access by 'any foreign national' to Mythos 5 and Fable 5, was issued on June 12th^[1]. Anthropic took its Mythos-class models offline immediately after, with cofounder Tom Brown later replacing CEO Dario Amodei in negotiations with the Trump administration. The US Department of Commerce had tested Fable 5 before its release and raised no complaints, according to The Verge^[1]. However, the administration reportedly acted after Amazon CEO Andy Jassy flagged a potential vulnerability in Fable 5. Anthropic has committed to working with the US government on protocols and standards for Mythos-class models^[2]. The company needs revenue from Mythos to pay for its recent compute deals, including a $15 billion per year agreement with SpaceX for access to its data centers^[1]. The approval of Mythos 5 for a select group of users is seen as a partial resolution, but Fable 5 remains in limbo with no apparent timeline for rollout. The Trump administration's decision has raised concerns about the US AI industry's competitiveness, with some experts warning that the government's actions could lead to other countries developing their own AI systems^[1].