Pentagon drops Anthropic, signs new AI deals with OpenAI and xAI

The Pentagon is actively developing tools to replace Anthropic’s artificial intelligence software, marking a shift in how the military procures and deploys AI. According to Bloomberg, this development follows the collapse of a $200 million contract between Anthropic and the Department of Defense—an agency known under the Trump administration as the Department of War. The breakdown of the deal occurred over the last several weeks after the two parties failed to reach an agreement regarding the military’s demand for unrestricted access to Anthropic’s AI models.

The core of the dispute centered on safety and usage limits. Anthropic attempted to secure a contractual clause that would explicitly prohibit the Pentagon from using its AI systems for the mass surveillance of Americans. The proposed clause also sought to prevent the deployment of autonomous weapons that can fire without human intervention. The Pentagon refused to agree to these restrictions, leading to an impasse and the ultimate termination of the $200 million contract.

In response to the breakdown, the Pentagon has diversified its AI partnerships. The Department of Defense has signed a new agreement with Elon Musk’s xAI to integrate its large language model, Grok, into classified systems. Simultaneously, OpenAI has entered into its own agreement with the Pentagon. These moves allow the military to transition away from Anthropic’s technology while continuing to build out its computational capabilities.

The fallout has also led to severe regulatory consequences for Anthropic. Defense Secretary Pete Hegseth designated the company as a “supply-chain risk,” a classification usually reserved for foreign adversaries. This designation effectively bars any company doing business with the Pentagon from working with Anthropic. Anthropic is currently challenging this “supply-chain risk” designation in court, seeking to overturn the restrictions.

Despite the ongoing litigation, the military is moving forward with its alternative AI infrastructure. Cameron Stanley, the chief digital and AI officer at the Pentagon, confirmed that the department is actively pursuing multiple large language models for its government-owned environments. “Engineering work has begun on these LLMs, and we expect to have them available for operational use very soon,” Stanley said.

Why it matters

The Pentagon’s pivot away from Anthropic signals a major shift in defense AI procurement, prioritizing control over specific vendor relationships and forcing a realignment among top AI labs seeking government contracts.