Pentagon Tests OpenAI and Google AI to Replace Anthropic's Claude

The Pentagon has begun formal testing of AI models from OpenAI and Google as it races to replace Anthropic's Claude across classified military systems. According to Bloomberg, tests kicked off at the start of March with 25 of the department's 'power users' — just three days after Defense Secretary Pete Hegseth designated Anthropic a supply‑chain risk and moved to end its role as an AI provider.

Emil Michael, the undersecretary of defense for research and engineering, told Bloomberg Television that talks with Anthropic remain suspended because of the company's legal challenge, and that the Pentagon is ready to transition to other vendors. He said he expects new model releases from competitors to deliver similar capabilities every month or two.

The dispute traces back to two explicit red lines Anthropic has held since its founding: no mass surveillance on Americans, and no fully autonomous weapons that target people without human input. When the Trump administration demanded Claude be available for 'all lawful purposes' without contractual guardrails, CBS News reported that Anthropic refused, and CEO Dario Amodei said the company saw 'no choice but to challenge it in court.'

The Pentagon's position, per a senior defense official speaking to:² 'The military will not allow a vendor to insert itself into the chain of command by restricting the lawful use of a critical capability and put our warfighters at risk.' The department argued that mass surveillance and autonomous weapons are already restricted by existing laws and policies, making Anthropic's contractual bans redundant.

Anthropic's Claude was deployed on the Pentagon's classified networks as the primary AI provider, deeply integrated into the Maven Smart System — a digital mission control platform used for classified operations, including recent strikes on Iran, according to two sources cited by Bloomberg. The full scope of Claude's deployment in AI targeting has not been publicly disclosed.

The transition won't be trivial. ³ quoted Joe Saunders, CEO of RunSafe Security, who warned: 'These models are embedded across workflows, security‑accredited environments, and mission‑specific processes. Even when other models are available, each one requires validation, and in many cases, re‑authorization before it can be used in operational settings.'

The Pentagon's CTO framed the Anthropic situation as a catalyst for a long‑overdue diversification strategy. Federal News Network reported Michael's comments at the McAleese defense conference:

The Pentagon-Anthropic split carries three lessons for builders targeting government contracts. First, multi‑model compatibility is now table stakes — the DoD explicitly stated that relying on one provider was a mistake. Second, guardrails have a real cost. Anthropic is potentially forfeiting billions in revenue by standing on principle, but the flip side is that builders who can offer flexible deployment models (with or without restrictions) gain a competitive edge. Third, the 180‑day transition timeline set a precedent — defense agencies now expect AI vendors to be replaceable within six months, which means API compatibility and standardized interfaces are becoming procurement requirements.

Bloomberg reported that early Pentagon tests show models respond differently to the same queries, and that varying prompts across models has helped optimize performance. The Pentagon may consider making the final evaluation public — which would give builders a rare window into how defense agencies actually benchmark AI tools.