A new artificial intelligence model, Mythos, developed by Anthropic, successfully identified vulnerabilities within highly sensitive US government computer systems during a testing exercise. The AI was able to penetrate nearly all classified systems in a matter of hours, according to a US Senator’s disclosure. The testing was conducted in collaboration with US intelligence agencies as part of Project Glasswing. This exercise aimed to assess the potential risks posed by advanced AI technologies. The findings highlight concerns about the security of classified information in the face of rapidly evolving AI capabilities. Anthropic and government agencies are now analyzing the results to strengthen defenses against potential threats. The incident underscores the need for proactive measures to safeguard sensitive data from AI-driven attacks.