DOD nears 90% cutover of classified AI off Anthropic
Maven warfighters say they prefer ChatGPT and Grok over what they left behind, and the multi-vendor shift was overdue regardless of the contract fight.
TL;DR
The Pentagon has migrated roughly 90% of its classified AI workloads away from Anthropic's Claude and is on track to complete the transition by the end of September, Under Secretary Emil Michael told reporters Thursday. Palantir's Maven Smart Systems cut over months ago. The move follows DOD's February designation of Anthropic as a national security supply-chain risk (a designation the company is challenging in two federal courts) after negotiations collapsed over Anthropic's insistence on ethical-use restrictions the department wouldn't accept. Meanwhile, the eight-vendor coalition DOD announced in May (OpenAI, Google, SpaceX, NVIDIA, Reflection, Microsoft, AWS, Oracle) is deploying frontier models across IL6 and IL7 classified networks.
Michael laid out the migration timeline at NDIA's Emerging Technologies for Defense Conference. "I'd say about 90% has transitioned. All of the Maven Smart Systems or Palantir work has been transitioned months ago, and we're on track to get it done by the end of the month," he said. The remaining 10% covers other classified systems where Claude was embedded, the department expects full cutover by October.
The multi-model approach now in place (spanning OpenAI's ChatGPT, xAI's Grok, Google's Gemini, and others across the eight-vendor agreement signed in May) was, in Michael's framing, long overdue regardless of the Anthropic dispute. "One of the mistakes, independent of whether we had a dispute with Anthropic, is the fact that the Pentagon was solely reliant on one model," he said. "So the number one idea is to get multiple models, because as we all know, every month this industry changes (maybe every week) who's leading and what."
Early feedback from operators supports the shift. Michael told reporters that Maven warfighters "have liked it a lot better than Anthropic" after switching to newer ChatGPT versions, and found Grok "way faster than both of them." Each model, he said, carries different attributes depending on the mission.
The legal fight continues on two tracks. Michael noted that Anthropic's challenge in the Northern District of California didn't yield a preliminary injunction blocking the supply-chain-risk designation. The second case, in the D.C. Circuit, is what Michael called "the more important one" (a forum he described as "more of a national security-oriented" venue) with a decision expected in the next month or two. Until then, the designation stands, and DOD isn't waiting.
Published ·Deep Fathom