The Department of Defense has entered into an agreement with Elon Musk’s artificial intelligence firm xAI to provide its Grok model for integration into classified military systems, a defense official confirmed.
With this arrangement, Grok will be the second AI model authorized for the military’s highest-sensitivity applications, spanning intelligence analysis, weapons development and battlefield operations. Until now, Anthropic’s Claude had been the only model accessible in those classified environments.
The Pentagon’s dialogue with Anthropic has intensified into a dispute centered on the company’s safeguards for Claude. Defense officials have asked Anthropic to permit Claude’s use for "all lawful purposes," a standard that would remove certain restrictions the company has put in place. Anthropic has resisted that request, stating it will block uses of Claude for "mass surveillance of Americans and the development of fully autonomous weapons."
According to the defense official, xAI accepted the Pentagon’s "all lawful use" condition. It remains unclear whether xAI’s Grok will fully supplant Anthropic’s Claude within classified systems or how long any transition might take.
Defense Secretary Pete Hegseth is scheduled to meet with Anthropic CEO Dario Amodei at the Pentagon. A defense official said Hegseth will press Amodei and deliver an ultimatum tied to the company’s stance on usage restrictions.
The Pentagon has warned Anthropic that failure to acquiesce could prompt the department to designate the company a "supply chain risk," among other potential penalties. Officials have not specified the full range of actions they might pursue, nor have they outlined a timeline for replacement should Anthropic’s position remain unchanged.
What this means
- Grok’s clearance for classified use expands the pool of AI models the military can deploy for sensitive work.
- The dispute highlights tension between defense requirements and company-imposed safeguards on AI usage.
- Outcomes could affect defense procurement decisions and the availability of AI tools for classified operations.
Given the current public details, the scope and timing of any operational changes remain uncertain. Officials are conducting direct, high-level discussions with Anthropic while also moving to onboard an alternative model that meets the Pentagon’s stated legal-use standard.