The saga continues — and it keeps getting stranger. On May 1st, Pentagon CTO Emil Michael said in a CNBC interview that Anthropic remains on the Department of Defense blacklist. At the same time, he emphasized that Mythos — Anthropic’s most powerful AI model — is a “separate national security moment” being handled “government-wide.”
The backstory
The Pentagon designated Anthropic as a “supply chain risk” after the two sides failed to agree on how Anthropic’s models could be used by the military. The sticking point: the DoD wanted unfettered access to all models for all lawful purposes. Anthropic insisted its technology shouldn’t be used for fully autonomous weapons or domestic mass surveillance.
Mythos as a special case
Emil Michael is deliberately treating Mythos differently from the rest of Anthropic’s product lineup. The reasoning makes sense: Mythos has particular capabilities for finding cyber vulnerabilities and patching them. In plain terms — it’s one of the best tools in the world for making software more secure. According to Axios, the NSA is already using the model, even though Anthropic officially remains on the Pentagon blacklist.
Signs of a thaw
There’s movement. Anthropic CEO Dario Amodei visited the White House earlier this month, and both sides called the discussion “productive.” Trump himself told CNBC that a deal between Anthropic and the Pentagon is “possible.” In diplomatic speak, that means negotiations are underway.
My take
The situation is paradoxical: Anthropic is officially blocked, but its most powerful model is being used by intelligence agencies. Michael’s framing of a “separate security moment” is an attempt to resolve this contradiction without losing face. For Anthropic, it’s a double-edged sword — on one hand, confirmation that Mythos is indispensable. On the other, the realization that you’re serving a customer who simultaneously has you on a blacklist.
Sources: