Anthropic's Mistral Discovers Security Vulnerabilities in U.S. Government Secret Systems... Spurring Regulatory Debate

Anthropic's Mistral identified critical security vulnerabilities in the U.S. government's highly secure computer systems within hours, prompting regulatory action but also intensifying industry pushback. Officials confirmed the AI model detected flaws during joint tests with U.S. intelligence agencies, though they emphasized the model cannot exploit vulnerabilities within that timeframe. The test was part of 'Project Glasswing,' a joint initiative to protect critical software from security threats. Senate Banking Committee member Mark Warner highlighted the model's ability to penetrate classified systems rapidly, citing intelligence sources. Despite NSA and Anthropic's refusal to comment, the U.S. government has restricted foreign access to Anthropic's latest AI models, following a presidential executive order. Anthropic rejected the restrictions as overly broad, while industry leaders urged the Trump administration to repeal the guidelines, arguing the model's capabilities are not exclusive.