The US Department of Defense is reportedly expanding its use of artificial intelligence by bringing Grok AI into its most sensitive classified systems, a move that marks a shift in military AI strategy and defense partnerships. According to a report by Axios, the Pentagon has signed an agreement allowing Grok AI, the conversational model developed by Elon Musk’s xAI, to operate alongside other AI systems in classified environments.

Previously, Anthropic’s Claude was the only AI model approved for use within the military’s top-secret networks, where tasks include intelligence analysis, weapons development and battlefield planning. Until now, models such as Grok AI, OpenAI’s ChatGPT and Google’s Gemini were limited to unclassified military systems.

Under the reported agreement, Grok AI has accepted the Department of Defense’s requirement that its technology be available for “all lawful purposes,” including highly sensitive applications, which Anthropic has resisted due to its safety guardrails against certain uses.

The development is part of a broader dispute between the Pentagon and Anthropic over the latter’s restrictions on using Claude for functions such as mass surveillance and fully autonomous weapons, leading to heated negotiations and a scheduled meeting between Defense Secretary Pete Hegseth and Anthropic CEO Dario Amodei.

While it remains unclear how quickly Grok AI might replace Claude in classified systems, defense officials acknowledge the technical complexity of transitioning such deeply embedded infrastructure. OpenAI and Google are also reportedly negotiating deals to bring their models into similar spaces, though neither has finalized an agreement as part of this initial expansion.

For now, Grok AI’s new role shows how military AI strategy is evolving under pressure to diversify its technological base and align capabilities with defense policy requirements.