Pentagon Deploys Anthropic's Mythos AI Despite Ongoing Phaseout

The United States Department of Defense is deploying Anthropic's unreleased AI cybersecurity model, Mythos, to identify and patch software vulnerabilities across government systems, even as it moves to phase out all Anthropic products in the coming months. The deployment falls under Project Glasswing, a controlled program launched on April 7 that grants select organizations access to the Claude Mythos Preview model for defensive cybersecurity purposes.

Pentagon Chief Technology Officer Emil Michael announced the deployment at a conference in Washington, calling it a national security moment. He acknowledged the dual-edged nature of the technology: AI enables faster vulnerability patching but also allows adversaries to exploit those same weaknesses more quickly.

The partnership comes against a backdrop of significant tension between the two parties. The Department of Defense previously designated Anthropic as a supply-chain risk, prompting the AI company to sue the Trump administration in March to overturn the blacklisting. That legal dispute remains unresolved even as the Pentagon now relies on Anthropic's most advanced unreleased model for critical defensive operations. The simultaneous phaseout plan and emergency deployment illustrate the competing pressures the Pentagon faces between managing perceived vendor risks and urgent cybersecurity needs.