Anthropic’s Mythos mess is only getting worse
TL;DR
Anthropic’s Mythos-class models have been offline for two weeks after the Trump administration’s June 12 order targeting foreign-national access to Mythos 5 and Fable 5. The Verge reports that high-level talks in Washington are still stuck, with Anthropic repeatedly saying there is no update to share. The fight centers on export controls, security fears, and a reported Fable 5 guardrail weakness that security expert Katie Moussouris argues has been overstated.
Nauti's Take
The problem is not that governments are asking hard security questions. The problem is that one of America’s most important AI companies appears trapped in an improvised political process with no clear threshold, timeline, or exit path.
If Washington wants to regulate frontier AI, it needs durable criteria, fast procedures, and auditable conditions. Otherwise security starts looking less like governance and more like unmanaged panic.
Briefingshow
This is bigger than Anthropic. If the US government can abruptly sideline frontier models without a clear rulebook, every major AI release becomes a political and regulatory risk. Customers, investors, and developers now have a new platform risk to price in: the model may work, but the approval environment may not.
Sources