Pentagon Tests Alternative AI Models as It Moves Away From Anthropic’s Claude
The U.S. Department of Defense is reportedly evaluating multiple artificial intelligence models as it looks for alternatives to Anthropic’s Claude, according to a Bloomberg report.
The testing program reportedly involves 25 high-level users within the Pentagon and forms part of a broader effort to diversify AI providers used across defense operations and classified systems.
Pentagon Began AI Evaluations After Anthropic Was Designated a Supply-Chain Risk
According to reports, testing of alternative AI models started in early March, only days after Defense Secretary Pete Hegseth classified Anthropic as a potential supply-chain risk.
The designation was reportedly linked to concerns surrounding the company’s stance on AI safety guardrails and technology restrictions. Following the decision, the Pentagon moved to remove Anthropic from its role as an AI tools provider.
Anthropic has challenged the classification in court, arguing the decision could lead to billions of dollars in lost revenue.
Anthropic’s Claude Previously Supported Classified Military Operations
The U.S. military has reportedly used Anthropic’s Claude within the Maven Smart System, a digital mission-control platform supporting classified operations, including missions related to Iran.
Claude is believed to have gained adoption among Pentagon personnel, although the full extent of its use in AI-assisted targeting and defense operations has not been publicly disclosed.
Defense Department Plans to Phase Out Anthropic Products
Following the separation from Anthropic, the Pentagon reportedly established a six-month timeline to reduce and eventually end the use of the company’s AI products.
More recently, the Department of Defense announced new agreements with multiple AI companies to deploy alternative models on classified government networks.
The move reflects a strategy aimed at avoiding dependence on a single AI provider while expanding access to competing technologies.
Pentagon Says Talks With Anthropic Remain Suspended
Emil Michael, U.S. Undersecretary of Defense for Research and Engineering, stated in an interview with Bloomberg Television that discussions with Anthropic remain paused due to the company’s legal challenge over the supply-chain risk designation.
Michael indicated the Pentagon is prepared to transition toward other AI vendors and expects competing models to achieve capabilities similar to Anthropic’s systems within short development cycles.
He also suggested that Anthropic’s ideological positions may not fully align with the Defense Department’s operational objectives, despite adoption of the company’s models by other government agencies.
OpenAI and Google Models Reportedly Under Review
According to Bloomberg, AI systems developed by OpenAI and Alphabet-owned Google are among the technologies currently being tested.
These evaluations are reportedly taking place on a separate digital platform outside of the Maven Smart System.
Early testing results suggest different AI models generate varying responses to identical prompts. Officials reportedly found that adjusting prompts across systems can significantly influence performance and outcomes.
The Pentagon has not released performance metrics or interim rankings but may reportedly make final evaluation results public in the future.
Growing Competition for Government AI Contracts
The Pentagon’s latest AI testing initiative highlights intensifying competition among major artificial intelligence companies seeking government and defense contracts.
As AI becomes increasingly integrated into national security, intelligence analysis, and military operations, access to large government deployments could become a major growth driver for leading AI developers.
