Report: Anthropic’s Mythos AI Breached Nearly Every NSA Classified Systems in Just Hours

Anthropic has found itself at the center of that debate after abruptly restricting access to its advanced Fable 5 and Mythos 5 models following intervention from U.S. authorities. What began as a security concern has now evolved into a much bigger conversation about AI governance, national security, and control over frontier technology.

The controversy stems from a U.S. government directive issued earlier this month that required Anthropic to limit access to Fable 5 and Mythos 5. Rather than attempting to enforce nationality-based restrictions across users, Anthropic ultimately disabled access more broadly, impacting researchers and partners outside the United States.

The restriction immediately became headline after reports surfaced referencing comments made during a Senate Intelligence Committee discussion. According to a report from The Economist, Senator Mark Warner cited information allegedly provided by NSA and Cyber Command leadership regarding the cybersecurity capabilities of Anthropic’s Mythos model (via Tom’s Hardware). However, many of the claims circulating online remain unverified, and independent confirmation has not emerged.

Per the report, the shutdown directive came after an explosive Senate Testimony, in which, he claimed that Mythos AI breached NSA and Cyber Command systems within hours. “Encryption was a potent technology, but narrow in its application. AI is far more powerful and versatile. On June 11th Mark Warner, the vice-chair of the Senate Intelligence Committee, said that General Joshua Rudd, who leads the National Security Agency and the Pentagon’s Cyber Command, had told him that Mythos “broke into almost all of our classified systems, not in weeks, but in hours,” reported The Economist.

A previous report also claimed that Amazon CEO reportedly also had the role in advising the U.S. government about Fable 5’s advanced vulnerability discovery capabilities, which prompted in restriction of foreign access. Not to forget, Microsoft has also restricted Claude Fable 5 use as Anthropic’s new requirements allow prompts and outputs to be retained for up to 30 days.

As frontier models continue advancing, the battle over who controls them may become just as important as the technology itself.