According to insiders, the U.S. Cybersecurity and Infrastructure Security Agency (CISA) is currently using the "Mythos" model developed by the AI startup Anthropic. The model is being used to conduct a comprehensive audit of the software code of the U.S. government to prevent potential security threats.

This secret work is handled directly by CISA's "Attack Surface Assessment" team, which specializes in digital security assessments for various government agencies. So far, the model has successfully identified a large number of serious vulnerabilities and defects that could be exploited by foreign intelligence agencies or cybercriminals.

The Turbulent and Competitive Path of Collaboration

In fact, the collaboration between Anthropic and the U.S. government has experienced a series of dramatic ups and downs over the past few months. In February of this year, the company was briefly placed on the "supply chain risk" blacklist by the Pentagon after refusing to remove the safety measures in its products that prohibit use in autonomous weapons.

Subsequently, Anthropic privately introduced the "Mythos" model, specifically designed to identify cybersecurity vulnerabilities, which quickly eased the tension between the two sides. The U.S. National Security Agency (NSA) had also expressed high praise for the model's deep reasoning and vulnerability detection capabilities during previous internal tests.

Global Access Restrictions Officially Lifted

In addition to a special version for government use, Anthropic previously launched a compliance version for the public called "Fable." However, shortly after the model was launched, the White House suddenly requested a complete ban on overseas users accessing the model, causing a global technological blockade.

Luckily, with the delicate balance achieved through the negotiation of regulatory and security boundaries, the U.S. government officially lifted the export restrictions on the model last week. Currently, major national security agencies are still conducting in-depth assessments of this AI technology to gain a technological advantage in digital warfare.