United States authorities are using Anthropic’s Mythos artificial intelligence model to help secure federal government software, according to a Reuters report citing three people familiar with the effort. The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has deployed Mythos to scan code repositories across multiple federal agencies in order to identify and remediate security vulnerabilities before they can be exploited by foreign intelligence services or criminal hackers. The initiative is led by CISA’s Attack Surface Evaluation team, a specialized group that conducts digital defense assessments and simulated hacking exercises across the federal landscape, and has already uncovered a large number of software flaws. United States cybersecurity officials have not disclosed which agencies or systems Mythos has examined, nor have they provided details on the severity or nature of the vulnerabilities identified so far. The use of Anthropic’s technology is notable because relations between the company and the U.S. government have been strained since Anthropic refused administration requests to weaken safeguards that prevent its models from being used for autonomous weapons or domestic surveillance, a stance that led the Pentagon to classify the firm as a supply-chain risk. Separately, the National Security Agency is also believed to be using Mythos, after a U.S. official told the Associated Press that one of Anthropic’s models recently found vulnerabilities in highly sensitive and secure government computer systems during a testing exercise.

Prepared by Jonathan Pierce and reviewed by editorial team.