Pentagon expands AI use with Google Gemini deal

The Pentagon has reached an agreement with Google to expand the use of its generative AI model Gemini in classified operations, the company told several U.S. media outlets, AFP reported Wednesday.

Like OpenAI, Google will now replace Anthropic, whose Claude model had until now been the only one authorized for classified operations, CE Report quotes AGERPRES.

At the end of February, the Trump administration decided to terminate all contracts with Anthropic, a move challenged in court by the California-based startup.

The administration later reached an agreement with OpenAI to integrate its models into classified operations, though the process is expected to take several months.

According to The Information, xAI, the AI startup founded by Elon Musk, has also secured a deal with the Pentagon following the split with Anthropic.

“Relying too heavily on a single provider is never a good thing,” Pentagon AI chief Cameron Stanley told CNBC.

According to multiple U.S. media reports, the agreement with Google, like those with OpenAI and xAI, allows the use of AI models within the limits of the law.

This provision gives the Department of Defense more flexibility than its previous contract with Anthropic, which excluded mass surveillance of the U.S. civilian population and lethal attacks.

On Monday, more than 600 Google employees signed a letter urging the company’s leadership to stop supplying its models to the U.S. military for classified operations.

In 2018, internal pressure led Google to withdraw from Project Maven, which used AI to analyze images collected by drones.