CISA Reportedly Uses Anthropic's Mythos to Scan Government Code Repositories

The U.S. Cybersecurity and Infrastructure Security Agency has previously struggled to obtain advanced artificial intelligence tools as it confronts increasingly complex cyber threats. With hostile actors beginning to use AI in attacks, protecting federal digital infrastructure from vulnerabilities has become critical. CISA's integration of Anthropic's Mythos AI model into its defenses marks a major shift in the U.S. government's cybersecurity strategy.

According to reports in July 2026, CISA's attack-surface assessment team has begun using Mythos to comprehensively scan government code repositories hosted on platforms including GitHub. People familiar with the matter said the audit program had identified numerous potential vulnerabilities and sharply accelerated defensive work. Cybersecurity professionals cautioned, however, that the model currently produces a high rate of false positives and that its automatically generated patches could destabilize systems.