The US Cybersecurity and Infrastructure Security Agency is using Anthropic’s AI model Mythos to scan government code repositories for software flaws, according to three people familiar with the work.

CISA’s Attack Surface Evaluation team is running the audits and conducts digital security assessments and hacking exercises across the government.

Two sources said the reviews have already found many vulnerabilities, although their number or severity could not be verified.

Public material describes Mythos as a cybersecurity model.

Anthropic said a preview version has discovered thousands of high-severity software flaws, while the UK AI Security Institute said a preview version could discover and exploit vulnerabilities autonomously.

Separate testing with US intelligence agencies also found that Mythos identified vulnerabilities in highly sensitive government systems within hours, suggesting CISA’s use of the model is not an isolated experiment.

The work also shows continued federal use of Anthropic’s tools after the Pentagon moved to blacklist the company in February, and a judge blocked that action in March.

🔗 Source: Reuters