Zerohedge: US Restricts AI Access for Foreign Nationals

Summary

Just four days after releasing its advanced "Mythos-class" AI model, Claude Fable 5, AI lab Anthropic has been compelled by the U.S. government to suspend access for all foreign nationals, including employees. This export control directive, issued by federal authorities citing national security, impacts Fable 5 and Mythos 5 models globally. Anthropic stated on X that this order was necessary to ensure compliance and was received around 5:21 p.m. ET, forcing an immediate shutdown for all customers. Access to other Claude models remains unaffected. The government's concern reportedly stems from potential methods of bypassing or "jailbreaking" Fable 5, despite Anthropic implementing robust safeguards and conducting thousands of hours of testing. While Anthropic believes its safeguards are highly effective and no universal jailbreaks have been found, they acknowledge that perfect resistance is likely impossible and have adopted a defense-in-depth strategy, including a 30-day customer data retention policy to monitor and mitigate attacks. This development follows positive analyst reviews of Fable 5, with one BMO analyst calling Anthropic a leading pure-play AI lab and highlighting its strengths in coding, agents, and enterprise applications.