Bond.az -- The Pentagon is testing AI models with 25 power users to find alternatives to Anthropic's Claude, Bloomberg reported.
Testing began in early March, three days after Defense Secretary Pete Hegseth designated Anthropic as a supply-chain risk. The company is challenging the designation in court.
The US military used Claude for the Maven Smart System targeting Iran. Claude had gained users, but its full deployment scope is unknown.
The Pentagon set a six-month timeline to phase out Anthropic products. It has signed new agreements with multiple AI vendors for classified networks.
Undersecretary Emil Michael said talks with Anthropic are suspended due to the legal challenge. He expects rival models to match Claude's capabilities every 1-2 months.
Models from OpenAI and Google are being tested on a separate platform. Early results show different responses to same queries, and fine-tuning prompts improves performance.
