The U.S. government ordered Anthropic to take its new Fable 5 AI model offline within 90 minutes after Amazon flagged a potential security flaw, raising questions about how the Trump administration will regulate AI companies going forward.

What happened: Anthropic released Fable 5 (a safeguarded version of its Mythos model) to the public last week. Mid-week, Amazon researchers discovered a potential jailbreak and alerted Anthropic. After Amazon CEO Andy Jassy contacted the Trump administration on Friday with concerns, the government ordered Anthropic to shut down both Fable 5 and the underlying Mythos model within 90 minutes. Anthropic complied and took both models offline, restricting access to foreign nationals and others. As of Tuesday, Fable 5 remains unavailable.

Why it matters: Anthropic itself has long argued that advanced AI could be dangerous and that government oversight was needed. Now that regulation is arriving, but through an emergency shutdown rather than a deliberate safety framework—raising doubts about whether the government's approach will be a serious regulatory regime or a tool for political leverage against companies. The incident sets a precedent for how AI models will be treated and suggests future releases may face similar government intervention.

What to watch: Independent security researchers had actually tested Fable 5's safeguards before the shutdown and found them solid, suggesting the Amazon jailbreak's severity remains disputed. The broader question is whether Fable comes back online this week or stays offline, and how that outcome shapes AI regulation going forward—a pattern likely to repeat with every new model release.