The Pentagon struck an agreement with Elon Musk’s xAI that could permit the Grok chatbot limited access to classified Department of Defense systems and is seeking a similar deal with Google for Gemini. Officials hope those agreements will pressure Anthropic to relax some restrictions on its Claude model, which is currently the only AI approved for classified environments. Grok’s history of producing sexually explicit and racist content has raised significant security and ethical concerns. High-stakes talks between Defense Secretary Pete Hegseth and Anthropic CEO Dario Amodei are expected to clarify whether Claude’s guardrails will be eased.
Pentagon Deal Could Put Elon Musk’s Grok Chatbot Inside Classified DOD Systems — Security, Ethics Concerns Raised
Harun Ozalp/Anadolu via Getty Images
The Pentagon has signed an agreement with Elon Musk’s AI company xAI that could allow the Grok chatbot limited access to classified Department of Defense systems, according to reporting by The New York Times. The Defense Department is also reportedly pursuing a similar arrangement with Google for its Gemini model.
What The Deal Could Mean
Officials say they hope the xAI and Google agreements will give the Pentagon leverage to press Anthropic — the company whose Claude model is currently the only AI cleared for use in the military’s classified environments — to relax some of Claude’s operational restrictions for lawful defense applications.
Elon Musk tried to laugh off the sexual imagery controversy surrounding Grok by asking it to put him in a bikini. / Leon Neal/Getty Images
Controversy Around Grok
Grok has a documented history of producing problematic outputs when used on X, the social network owned by Musk. Reports say the model has been used to generate sexually explicit images of women and minors and has produced racist commentary and praise of Adolf Hitler. French prosecutors reportedly raided X’s Paris offices as part of an inquiry into allegations that Grok was used to generate sexual images of minors.
“Anthropic knows this is not a get-to-know-you meeting. This is not a friendly meeting. This is a s--t-or-get-off-the-pot meeting,” a senior Defense official told Axios, describing upcoming talks between Defense Secretary Pete Hegseth and Anthropic CEO Dario Amodei.
Anthropic’s Position
Anthropic has said it may be willing to ease some restrictions but insists that guardrails remain to prevent mass surveillance of Americans and to avoid enabling weapon systems that operate without human oversight. Pentagon officials acknowledge Claude is currently regarded as more advanced and reliable for sensitive defense and intelligence tasks than Grok.
Defense Secretary Pete Hegseth summoned Anthropic CEO Dario Amodei to the Pentagon for a potentially tense meeting. / AAron Ontiveroz/The Denver Post
Politics And Regulation
Separately, the White House has reportedly urged Florida lawmakers to oppose a proposed state-level Artificial Intelligence Bill of Rights that would impose stricter limits on certain AI uses, including generation of explicit images involving minors. Administration officials say they prefer a single federal approach rather than a patchwork of state regulations; a White House official described claims that the administration opposes protections for minors as "fatally flawed." The Daily Beast has contacted the Pentagon and xAI for comment.
What’s Next: Defense Secretary Pete Hegseth is scheduled to meet Anthropic CEO Dario Amodei in high-stakes talks that could determine whether Claude’s guardrails are loosened and whether other models like Grok gain a foothold in classified military systems.
Help us improve.
