Anthropic's Mythos AI model found vulnerabilities in classified US government systems

Associated Press—An official confirmed to the Associated Press that Anthropic's Mythos AI model discovered security vulnerabilities in classified US government systems. The disclosure, reported simultaneously by AP and Reuters, marks one of the first public acknowledgments that a commercial AI model has been used against real classified infrastructure. CNBC noted the finding raises immediate questions about AI security testing protocols in government environments.

• Associated Press — US official confirmed AI model's access to classified systems to AP
• NDTV — Anthropic's Mythos Model Found Vulnerabilities In Classified US Government Systems
• CNBC — CNBC: finding raises questions about AI security-testing protocols in government
• Reuters — Reuters confirmed: Mythos found vulnerabilities in classified US systems