Pentagon Expands AI Model Testing as It Seeks Alternatives to Anthropic’s Claude

The US Department of Defense is accelerating its evaluation of artificial intelligence models as it searches for alternatives to Anthropic PBC’s Claude AI system, according to a Bloomberg report published Thursday. The Pentagon has reportedly begun testing several competing AI models with a group of 25 internal power users as part of a broader effort to diversify its AI suppliers for classified military operations.

The testing program reportedly started in early March, shortly after Defense Secretary Pete Hegseth labeled Anthropic a potential supply-chain risk. The designation was linked to concerns over the company’s stance on AI technology guardrails and national security policies. Following the decision, the Pentagon moved to phase out Anthropic’s role as an AI tools provider within six months.

Anthropic is currently challenging the designation in court, arguing that the decision could cost the company billions of dollars in future revenue. Despite the dispute, Claude AI remains integrated into the military’s Maven Smart System, a digital mission-control platform reportedly used for classified operations involving Iran.

The Pentagon is now reviewing alternative AI models from major technology firms, including OpenAI and Alphabet’s Google. According to the report, these models are being tested in a separate secure digital environment outside the Maven platform. Early evaluations suggest that each AI model produces different responses to identical prompts, prompting defense officials to experiment with prompt optimization strategies to improve performance.

US Undersecretary of Defense for Research and Engineering Emil Michael stated in a Bloomberg Television interview that discussions with Anthropic remain paused because of the ongoing legal challenge. He added that the Pentagon expects competing AI developers to release increasingly capable models every one to two months, potentially matching Claude’s performance.

The Defense Department recently announced new partnerships with multiple AI companies as it strengthens its classified AI infrastructure and reduces reliance on a single provider.