Government Shuts Down Anthropic's Most Powerful AI Models Over Security Breach

Amazon CEO Triggers Unprecedented AI Shutdown

In an unprecedented move, the U.S. government ordered Anthropic to disable its two most powerful artificial intelligence models on Friday after Amazon CEO Andy Jassy contacted senior White House officials about a security vulnerability . The shutdown of Fable 5 and Mythos 5 marks the first time a leading AI company has taken a publicly deployed model offline due to intervention from the federal government .

Amazon researchers discovered they could jailbreak Fable 5 to extract sensitive information useful for cyberattacks , prompting Jassy's late-night call to Treasury Secretary Scott Bessent and other senior officials . The Trump Administration cited national security concerns when it issued the export-control directive , requiring Anthropic to suspend access to the models for any foreign national .

Rather than attempt to selectively filter users, Anthropic chose to disable both models for all customers globally to ensure compliance . The company received the directive at 5:21 PM Eastern Time on Friday , giving it less than 90 minutes to implement the shutdown.

The Vulnerability That Sparked Crisis

Anthropic had released Fable 5 just days earlier, claiming it represented a new "Mythos-class" level of capability that exceeded any model the company had previously made available, particularly in identifying software vulnerabilities . Amazon's research showed that through a series of prompts, Fable 5 could be made to serve up information that could be used in cyberattacks .

White House AI adviser David Sacks said the administration warned Anthropic CEO Dario Amodei about the jailbreak and asked him to fix the flaw or pull the model, but Amodei refused . Sacks said the administration issued the export control "reluctantly" after that refusal . Adding to concerns, the White House suspected a China-linked group had accessed Mythos , raising fears about reverse-engineering.

Anthropic pushed back, arguing the jailbreak exposed only "a small number of previously known, minor vulnerabilities" and that similar capabilities exist in other publicly accessible AI models . The company stated it disagreed "that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people" .

Industry-Wide Implications

Anthropic warned that if this standard were applied industry-wide, "it would essentially halt all new model deployments for all frontier model providers" . The company defended its "defense in depth strategy," noting that "perfect jailbreak resistance is not currently possible" and that it had implemented monitoring systems to detect attacks.

The shutdown created immediate operational challenges, as many of Anthropic's own foreign-born engineers and researchers were barred from accessing their company's models . Enterprise customers including banks and government agencies that had been using the models for complex reasoning and vulnerability discovery were also affected .

This incident highlights the growing tension between AI innovation and national security concerns as models become increasingly powerful. Officials now consider this "the scary phase of AI" where models are deemed so powerful that full public release could unleash catastrophe . The precedent set by this government intervention will likely shape how future AI deployments are handled, as companies must now navigate not just technical challenges but also unprecedented regulatory scrutiny over their most advanced capabilities.