Anthropic's Mythos cybersecurity AI reaches NSA and Commerce Department, but America's main cyber agency CISA is left out

Anthropic released Mythos Preview, an AI model designed to find and patch security vulnerabilities in software. The NSA, Commerce Department, and other federal agencies are now using it, but the Cybersecurity and Infrastructure Security Agency (CISA)—the government's central coordinator for national cybersecurity—reportedly has no access to the tool.

Mythos works by analyzing code to automatically detect security weaknesses (like SQL injection flaws or unpatched exploits) that human engineers would normally spend weeks hunting down manually. This means security teams can patch vulnerabilities faster before attackers find them.

For federal agencies and private companies that rely on government cybersecurity guidance, this creates an awkward gap: CISA won't be able to test Mythos internally or advise the public on its security risks and effectiveness—potentially slowing adoption of a tool that could strengthen America's defenses. The Trump administration is reportedly negotiating broader access, suggesting this exclusion may be temporary.