Anthropic's Claude Fable 5, arguably the most capable AI model available to the public, was shut down for 19 days by the US government and just came back online today. On June 9 Anthropic launched Fable 5. Three days later the Commerce Department hit it with emergency export controls after Amazon researchers found a jailbreak that bypassed its safety filters to identify software vulnerabilities and write exploit code. The order banned any foreign national from using the model, including Anthropic's own non-citizen employees. Since they couldn't verify nationality in real time, they shut it down for everyone. To get it back online Anthropic trained a new safety classifier that blocks the technique in 99% of cases, agreed to give the government pre-release access to future models, and committed to real-time information sharing on jailbreaks. Commerce Secretary Lutnick signed off on the reversal June 30. The bigger story isn't the resolution. It's that when Washington wanted to act on a frontier AI model, it had no regulatory framework so it reached for export controls, a tool designed for missiles and weapons, and applied it to software with 72 hours notice. The government can and will ground the most capable AI models on the planet with almost no process, and for an industry building on top of these models, that risk hasn't been priced in.