Anthropic released and then pulled two frontier AI models, Fable 5 and Mythos 5, after a U.S. export-control notice flagged national security concerns. Amazon escalated the issue before the models were withdrawn. A reported jailbreak of the models was also involved in the sequence of events.
This episode matters because it is one of the first documented cases of the U.S. government directly intervening to restrict public access to a frontier AI release. The mechanics are specific: export controls, a cloud infrastructure partner sounding the alarm, and a security vulnerability all converging at once. That combination is new.
The full episode runs through the types of jailbreaks involved, the logic frontier labs use when weighing model releases, and the political architecture forming around AI access. One host, Josh, works with Anthropic as a contractor, which adds direct context worth hearing firsthand.
[WATCH ON YOUTUBE →]