Washington's new gate for frontier AI: a vetting order, and the first export controls on a commercial AI model
A June executive order lets US national security agencies test the most capable AI models for up to 30 days before release, on a voluntary basis. Ten days later, export controls took Anthropic's Claude Fable 5 and Mythos 5 offline worldwide. What each side said, how access came back, and what it means for users outside the United States.