Skip to main content

White House Reverses Course: CISA Deploys AI to Hunt Code Flaws

In a striking policy reversal, the White House has lifted a ban on a powerful AI model developed by Anthropic, clearing the way for the U.S. Cybersecurity and Infrastructure Security Agency (CISA) to deploy it for auditing government software code. The model, known internally as "Mythos," is now being used to hunt for security flaws that could be exploited by foreign adversaries or cybercriminals.

A Rocky Road to Collaboration

The partnership between Anthropic and the U.S. government has been anything but smooth. In February, the Pentagon placed Anthropic on a "supply chain risk" blacklist after the company refused to remove safety measures that prohibited the use of its AI in autonomous weapons. The move threatened to derail any government collaboration.

But behind the scenes, Anthropic had been developing a specialized version of its AI—Mythos—designed specifically for cybersecurity vulnerability detection. The model's capabilities quickly caught the attention of the National Security Agency (NSA), which praised its deep reasoning and ability to identify subtle code flaws during internal tests. That endorsement helped ease tensions and paved the way for a renewed partnership.

CISA's Secret Audit

CISA's "Attack Surface Assessment" team is now using Mythos to conduct comprehensive audits of software code across multiple government agencies. According to insiders, the model has already identified a significant number of serious vulnerabilities that could have been exploited by foreign intelligence agencies or cybercriminals.

"The model is like a tireless code reviewer that never misses a beat," one official said. "It's already found issues that human auditors might have overlooked."

Global Access Restrictions Lifted

Anthropic had previously released a public version of the model called "Fable," but the White House abruptly blocked overseas access shortly after its launch, citing national security concerns. The restriction effectively created a global technology blockade, drawing criticism from international partners.

Last week, however, the White House officially lifted the export restrictions, allowing global access to the model. The decision came after delicate negotiations over regulatory and security boundaries. Major national security agencies are now conducting in-depth assessments of the AI technology, hoping to gain a technological edge in digital warfare.

Key Points

  • The White House reversed a ban on Anthropic's "Mythos" AI model, allowing CISA to use it for government code audits.
  • The model has already uncovered critical vulnerabilities that could be exploited by foreign adversaries.
  • The collaboration faced earlier setbacks when Anthropic was blacklisted for refusing to remove safety restrictions on autonomous weapons.
  • The NSA's endorsement of Mythos helped ease tensions and led to the current partnership.
  • The White House lifted global access restrictions on the model last week, ending a months-long technology blockade.