Anthropic’s testing reveals classified U.S. government systems have vulnerabilities, says anonymous source

An AI model from Anthropic discovered vulnerabilities in top-secret U.S. government systems during a joint security test with intelligence agencies. This finding raises serious concerns about the robustness of government cybersecurity and highlights the potential risks posed by advanced AI tools. The implications could lead to a reevaluation of how classified systems are protected against both internal and external cyber threats.