Pentagon Deploys Anthropic’s Mythos Even as Phase-Out Clock Runs

The Department of Defense is simultaneously deploying Anthropic's restricted Claude Mythos Preview model to scan government networks for software vulnerabilities and executing a court-ordered phase-out of all Anthropic products, a posture that under secretary Emil Michael, the Pentagon's chief technology officer and under secretary of Defense for research and engineering, publicly acknowledged at POLITICO's National Security Summit on Tuesday. Speaking at a conference in Washington, D.C., Michael described the underlying challenge as a "national security moment." The dual track, deploying a tool from a vendor the Pentagon has labeled a national security threat while simultaneously litigating that very designation, defines the institutional tension now shaping federal AI procurement policy.

Anthropic announced Mythos earlier this year as a frontier model designed for cybersecurity tasks, but restricted public access because of the model's capacity for offensive cyberattacks. The company limited access to approximately 40 organizations, publicly naming only a dozen. Distribution runs through a program Anthropic calls Project Glasswing, announced on April 7, 2026, which allows select organizations to access the Claude Mythos Preview model specifically for defensive cybersecurity purposes. According to Anthropic, Mythos can detect decades-old vulnerabilities in web browsers, infrastructure, and software. Independent benchmarking adds technical weight to those claims: evaluations by the United Kingdom's AI Security Institute and Palo Alto Networks found that Mythos and OpenAI's GPT-5.5 have substantially outpaced prior trends in autonomous cyber capability, with the U.K.'s AI Security Institute reporting that Mythos completed a 32-step simulated attack in 6 of 10 attempts and GPT-5.5 in 3 of 10.

Reuters reports the Pentagon is using Mythos under Anthropic's Project Glasswing to patch vulnerabilities while also planning to wean federal systems off Anthropic products. Michael confirmed at the summit that the Pentagon intends to continue the phase-out while exploiting Mythos's defensive utility in the interim, and expressed confidence that competing models from other vendors could eventually fill the gap [POLITICO]. He suggested Anthropic's lead may be temporary, saying competing models from rivals including OpenAI, xAI, and Google are expected to develop similar capabilities soon. OpenAI's GPT-5.5-Cyber has been cited alongside Mythos as a parallel instrument for government vulnerability scanning [POLITICO]. The NSA's involvement extends the footprint further: the National Security Agency is using Anthropic's Mythos Preview, despite top officials at the Department of Defense, which oversees the NSA, insisting the company is a "supply chain risk," according to two sources cited by Axios. The U.K.'s AI Security Institute has also confirmed it has access to Mythos.

The legal backdrop frames the deployment's significance. On Feb. 27, 2026, President Donald Trump directed all federal agencies to cease using Anthropic's AI technology, and Defense Secretary Pete Hegseth designated Anthropic a "supply chain risk," a designation that followed weeks of failed negotiations over the military's use of Anthropic's Claude model. The authority invoked derives from two statutes: the Federal Acquisition Supply Chain Security Act (FASCSA, 41 U.S.C. §§ 1321–1328 and 4713) and 10 U.S.C. § 3252, which allow the government to designate a vendor as a supply chain risk and exclude it from government contracts. Anthropic is the first American company to receive the designation, which has historically been reserved for foreign adversaries. On March 9, 2026, Anthropic filed lawsuits in two federal courts challenging the designation. Because 10 U.S.C. § 3252 and FASCSA are subject to different legal authorities and judicial review processes, Anthropic filed actions in both the Northern District of California and the Court of Appeals for the D.C. Circuit.

The litigation has produced split rulings. A federal appeals court in Washington, D.C. denied Anthropic's request to temporarily block the Pentagon's blacklisting, while a judge in San Francisco federal court granted Anthropic a preliminary injunction that bars the Trump administration from enforcing the ban on the use of its Claude model. With the split decisions, Anthropic is excluded from DoD contracts but is able to continue working with other government agencies while litigation proceeds. The Northern District ruling was notably pointed: U.S. District Judge Rita Lin wrote that "nothing in the governing statute supports the Orwellian notion that an American company may be branded a potential adversary and saboteur of the U.S. for expressing disagreement with the government." She ruled that the government's actions violated Anthropic's First Amendment and due process rights.

The diplomatic track has also shown movement. Anthropic CEO Dario Amodei met with White House chief of staff Susie Wiles and Treasury Secretary Scott Bessent to discuss the use of Mythos within government and Anthropic's broader security practices; next steps after that meeting were expected to focus on how departments other than the Pentagon engage with the model, and both sides described the meeting as productive. The core contractual dispute that triggered the phase-out order centered on competing interpretations of permissible use: the Pentagon issued the supply chain risk designation after negotiations broke down over two red lines Anthropic insisted upon, namely that its AI tool not be used for mass surveillance of U.S. citizens and that it not be used for autonomous weapons. The Pentagon countered that it requires access to Anthropic's AI "for all lawful purposes," asserting that it could not allow a private company to dictate how it uses its tools in a national security emergency. No resolution has been announced, and the merits of both suits remain pending.

—