US Government Restricts Access to Anthropic AI Model After It Uncovers Classified System Vulnerabilities, Prompting Lawsuit

Key Intelligence

• ■The U.S. government has restricted access to Anthropic's advanced AI model, 'Mythos' (also referred to as 'Fable 5' by some users), for certain entities, including its own intelligence agencies like the NSA, due to export controls.
• ■The decision followed a test where the Anthropic AI model rapidly identified significant vulnerabilities within classified U.S. government systems within hours.
• ■A U.S.-based legal tech firm, Legion (parent company Fable AI), is suing the U.S. government, alleging that these restrictions unlawfully cut off its access to Anthropic's top-tier models, which it was using for its own product development.
• ■This incident highlights the complex challenges of regulating powerful AI technologies, balancing national security concerns with innovation, and has sparked debate on U.S. AI policy and its international implications, particularly concerning AI rivalry with China.