How the world's top AI models were revived
TL;DR
Anthropic’s new Mythos and Fable models were offline for about 20 days after Amazon flagged a possible jailbreaking flaw to the U.S. government, Axios reports. The Trump administration responded with broad export controls. Commerce Secretary Howard Lutnick reportedly told CEO Dario Amodei the models had to be fixed quickly or stay offline. Anthropic sent engineers to Washington. The NSA and the Center for AI Standards and Innovation reviewed fixes before the models were released again on July 1.
Nauti's Take
This is not a clean win for AI safety; it is a warning sign for messy governance. If Amazon raises the alarm, Washington spins up multiple agencies, and outside experts later say the issue is not unique to Anthropic, the missing piece is a credible, transparent review process.
Otherwise the next frontier model will not be judged only on technical risk, but also on who reaches the right official first.
Briefingshow
The episode shows how quickly one safety warning can turn into a geopolitical brake on frontier models. For labs, cloud partners and users, model releases are becoming less like normal product updates and more like regulatory processes involving agencies, diplomacy and export controls. The open question is whether this becomes a clear standard or just another ad hoc power play.
Sources