Amazon CEO Andy Jassy personally flagged security vulnerabilities in Anthropic’s newest AI models to senior Trump administration officials, triggering the first-ever US export controls on commercial AI technology. Anthropic was forced to suspend global access to Claude Fable 5 and Mythos 5 barely days after their June 9 launch.
The internal Amazon team reportedly jailbroke Claude Fable 5 within days, extracting information useful for cyberattacks. Jassy escalated the findings directly to officials, including Treasury Secretary Scott Bessent, by June 12. The government immediately barred foreign nationals from accessing both models.
The restrictions applied globally, even locking out non-US Anthropic staff in London and Tokyo. Amazon, Anthropic’s largest backer with tens of billions invested, bypassed private channels to go directly to the federal government. Anthropic considers the export controls a disproportionate regulatory overreach.
This marks a shift in AI export policy, which previously targeted hardware chips, not software models accessible via API. Experts debate whether denying allied researchers access will ultimately hinder vulnerability fixes and fragment the global ecosystem.