Anthropic’s Mythos mess is only getting worse
TL;DR
Anthropic’s Mythos 5 is partly back after two weeks of talks with the Trump administration, but access is limited to selected organizations such as cyber defenders, infrastructure providers and US government users. Fable 5, the public-facing Mythos-class model, is still stuck. The Verge reports no clear timeline for a broader rollout deal. The June 12 export-control order remains in place. The government carved out an exception for Mythos 5 and reserved the right to adjust the license scope later.
Nauti's Take
The Mythos case looks like regulation by panic button. Cyber-capable models need limits, but a black-box approval process for selected partners is not a serious framework.
For AI vendors, this is the warning shot: selling raw model power is no longer enough. Trust now also depends on governance, export strategy and the ability to avoid cutting customers off in the middle of real work.
Briefingshow
This is bigger than Anthropic. If the US government can halt frontier models on short notice and reopen them only for approved users, AI release planning becomes political as much as technical. For customers, the best model matters less if access, compliance and nationality rules can change overnight.
Sources