ANTHROPIC’S HACKING AI GETS A CLASSIFIED DEMO IN CONGRESS — LAWMAKERS SAW WHAT IT FOUND IN EVERY MAJOR OPERATING SYSTEM Anthropic brought its Mythos artificial intelligence model to Capitol Hill this week for a closed session with the House Homeland Security Committee. Lawmakers did not just hear a briefing. They watched a live demonstration. Anthropic’s red team walked members of Congress through exactly how Mythos identifies and exploits software vulnerabilities, including flaws it has already found in every major operating system and web browser in active use today. Mythos was never given a full public release. Anthropic held it back specifically because of what it can do to security infrastructure. The model has identified thousands of high-severity vulnerabilities, including gaps that would give an attacker deep access to enterprise and government systems. The NSA is among the agencies already using it, a detail confirmed by multiple sources despite the Defense Department having placed Anthropic on a supply chain risk list, a bureaucratic contradiction that lawmakers pressed the company on directly. Members asked pointed questions about who has access to Mythos, what oversight exists, and how the federal government intends to manage a piece of technology this dangerous in the wrong hands. They also asked whether Anthropic’s legal dispute with the Pentagon is slowing AI integration at agencies like CISA. The session signals how fast the national security conversation around AI has shifted. Theoretical dangers are now real products sitting in government hands, finding real holes in real systems. Congress is only beginning to understand what that means. Keywords: Anthropic Mythos Congress briefing, AI hacking vulnerabilities, House Homeland Security AI, Mythos cybersecurity national security