The US Final AI Regulation Leaves a Loophole, Allowing Companies Like Cisco to Access Anthropic's Most Powerful Secure Model

According to Insight One monitoring, the U.S. government's export control order regarding Anthropic has failed to completely block access. Cybersecurity firms Dragos and Cisco Systems have both confirmed that they retain access to the Mythos beta version.

The Mythos beta version has demonstrated strong vulnerability detection capabilities during testing, able to find thousands of security flaws in mainstream operating systems and browsers. To prevent misuse, Anthropic restricted access to the Mythos beta version to the Project Glasswing alliance internally in April. The Project Glasswing alliance consists of approximately 200 organizations, including banks and tech companies. As an alternative, Anthropic released the Fable 5 model to the public, which has had its cybersecurity task execution permission removed.

Following the issuance of export license review requirements, the U.S. government ordered Anthropic to block access to Fable 5 and Mythos 5 to prevent unauthorized foreign nationals from using them. Anthropic subsequently shut down public access to Fable 5 and paused the distribution of Mythos 5 to Project Glasswing members. However, the U.S. government directive did not mention the Mythos beta version. Anthropic stated in an official blog post that the "remaining models" are unaffected but did not explicitly mention the Mythos beta version.