Anthropic’s Mythos model exposes major flaws in classified US systems
An official confirmed Anthropic’s Mythos model found vulnerabilities in secure US systems within hours during Project Glasswing testing. The Trump administration subsequently issued restrictions on Mythos 5 and Fable 5, forcing a global service halt.
Summary:
-
Anthropic’s Mythos model identified vulnerabilities in secure US government computer systems within hours during an authorized test called Project Glasswing, according to a CNBC report citing AP.
-
Senator Mark Warner disclosed that the tool successfully broke into almost all classified systems, quoting NSA and US Cyber Command chief Gen. Joshua Rudd.
-
The Trump administration issued a directive restricting foreign nationals from accessing the latest Fable 5 and Mythos 5 models.
-
Anthropic disabled the affected models for all global customers to comply with the directive, sparking backlash from over 100 cybersecurity leaders who warn the move harms American cyber defense.
A profound tension has emerged between the tech sector and Washington following revelations that Anthropic’s advanced artificial intelligence model exposed widespread vulnerabilities in classified US government systems. During an authorized testing exercise known as Project Glasswing, the company teamed up with US intelligence agencies to evaluate national security infrastructure. According to a CNBC report citing AP, the company’s Mythos model successfully identified critical flaws within hours of being deployed, though officials noted this does not mean the model actively exploited them.
The scale of the model’s capabilities became public when Senator Mark Warner mentioned the exercise during a Senate committee hearing, noting that the tool broke into almost all classified systems in hours rather than weeks. This assessment was attributed directly to General Joshua Rudd, head of the National Security Agency and US Cyber Command. In response to the perceived national security risks, the Trump administration issued a strict directive restricting foreign nationals from using Anthropic’s newest models, Mythos 5 and Fable 5.
To ensure compliance with the sudden government mandate, Anthropic completely disabled access to these models for all of its customers globally. This dramatic move has drawn sharp criticism from industry leaders at firms like Nvidia and Adobe, who argue that restricting these defensive benchmarking tools ultimately handicaps American cybersecurity against rapidly advancing foreign adversaries.
—
The immediate suspension of Anthropic’s flagship models injects severe uncertainty into the commercial enterprise AI market, disrupting organizations relying on advanced models for automated security auditing. By forcing a total service halt to comply with foreign national restrictions, the government directive sets a volatile precedent for regulatory intervention. This friction threatens to stall private-sector AI adoption and could pivot commercial market demand toward open-source alternatives or competitors unaffected by the specific mandate.
This article was written by Eamonn Sheridan at investinglive.com.