Anthropic’s Mythos AI Found Flaws In Classified US Systems Within Hours

Anthropic’s Mythos model reportedly identified vulnerabilities in highly sensitive U.S. government systems during a restricted testing program, raising fresh questions about AI’s (artificial intelligence) role in national security and the company’s strained ties with Washington.
 

TL;DR

• Anthropic’s Mythos AI model found vulnerabilities in classified U.S. government systems during Project Glasswing.
• Senator Mark Warner said the model “broke into almost all” classified systems within hours, though the AP clarified discovery did not mean exploitation.
• The report comes amid U.S. export restrictions and a dispute over government access to Anthropic’s AI models.

Mythos AI Testing Raises National Security Questions

Anthropic’s Mythos model identified vulnerabilities in highly sensitive U.S. government computer systems during a testing exercise, according to the Associated Press.

The test was conducted under Project Glasswing, a restricted program involving Anthropic and Washington’s intelligence agencies. The goal was to find and fix weaknesses in critical software before attackers could exploit them.

The development highlights how advanced AI models are being tested not just for productivity or coding, but also for defensive cybersecurity use cases in sensitive government environments.
 

Senator Mark Warner Says Mythos Moved Fast Across Classified Systems

The testing surfaced publicly after Senator Mark Warner of Virginia referred to it during a congressional hearing this month.

Warner said he had been told by National Security Agency chief Joshua Rudd that Mythos “broke into almost all of our classified systems, not in weeks, but in hours.”

However, the AP, citing an unidentified U.S. official, reported that while Mythos identified some vulnerabilities within hours, it did not necessarily exploit them in that timeframe. That distinction matters because vulnerability discovery and active compromise are very different levels of security impact.
 

Anthropic’s U.S. Government Relationship Faces Fresh Strain

The report arrives as Anthropic’s relationship with the U.S. government remains tense. The IPO-bound AI company had refused to allow the U.S. military to use its models for domestic surveillance and fully autonomous weapons systems.

The government later placed Anthropic on a national security blacklist.

This month, the U.S. government also ordered Anthropic to suspend exports of its latest Mythos and Fable AI models to destinations worldwide and to all foreign nationals, citing national security concerns.

The New York Times also reported that the NSA lost access to Mythos amid the dispute.

Neither the White House, Anthropic, nor the Department of Defense immediately responded to Reuters’ requests for comment.