Anthropic's Top AI Scientist Locked Out of His Own Company's Most Powerful AI Because He's Not American

Anthropic abruptly disabled its two most advanced AI models, Claude Fable 5 and Mythos 5, on June 10 after the U.S. Commerce Department issued an export control directive. The order, delivered in a letter to CEO Dario Amodei, required compliance within 90 minutes, forcing the company to shut down access worldwide for all users. While the directive targeted foreign nationals, Andrej Karpathy—a top AI scientist with a U.S. green card—was incorrectly impacted, sparking controversy over the policy’s execution. The shutdown followed a report from Amazon, Anthropic’s largest investor, claiming researchers had extracted restricted cybersecurity information from Fable 5 using prompts. Amazon CEO Andy Jassy alerted Treasury Secretary Scott Bessent, who issued the directive without detailing specific security risks. Anthropic disputes the findings, arguing the vulnerabilities were minor and already known, with CEO Katie Moussouris of Luta Security calling the shutdown an ‘own goal.’ Karpathy, a prominent AI researcher who previously led teams at Tesla and OpenAI, joined Anthropic last month to head pretraining research. Despite his green card status, he was temporarily locked out, becoming a symbol of broader concerns about U.S. AI regulations. Other non-U.S. citizens, including co-founder Chris Olah and philosopher Amanda Askell, remain affected, though Anthropic declined to confirm individual cases. The incident coincides with Anthropic’s confidential IPO filing, valuing the company near $965 billion, though pre-IPO futures dropped 3% following the news. The Pentagon also cited the shutdown as evidence of Anthropic’s exclusion from military contracts, further escalating tensions over AI governance. For users, the move means hundreds of millions of customers lost access to advanced models, raising questions about the balance between security and innovation in AI development.