Primakov/Shutterstock
xAI’s Grok, which is facing bans in some regions, might soon have access to classified military intelligence, as the Pentagon looks for AI chatbots to replace Anthropic’s Claude. As reported by Axios, Anthropic refuses to allow its chatbot to be used to assist with the development of fully autonomous weapons or the deployment of mass surveillance tools. As a result, the Pentagon has been in search of more willing partners, and seems to have found one in xAI.
While it’s unclear if Grok will be able to effectively replace Claude or how long it will take for the Pentagon to integrate the AI into its systems, Axios reports that the Pentagon is also in talks with other AI chatbot partners. For example, AI chatbots can currently be used in unclassified systems, like OpenAI’s ChatGPT and Google’s Gemini, and the U.S. Department of Defense continues to engage in talks with those players as potential replacements for Claude.
Grok will have access to classified information
bella1105/Shutterstock
While xAI made a government-specific version of Grok in 2025 for federal agencies to use, bringing Grok to classified systems in the military will be much more serious and rigorous. After all, the AI chatbot needs to meet security standards, while everything would be deployed locally. That said, if xAI becomes the second AI chatbot to actually be used by the Pentagon for classified documents, it wouldn’t mean Elon Musk or xAI would have access to sensitive data.
While xAI is currently leading the pack in the race to replace Anthropic’s Claude, Axios reports that the Pentagon is having conversations with a “sense of urgency” with OpenAI as well. A separate New York Times report says that the DoD is closer to a deal with Google than OpenAI. Depending on how these discussions go, the Pentagon could even go with all three of the other AI chatbots, as long as they meet the criteria to handle classified information.
For now, though, xAI is the only confirmed partner. Given that we’re just a few months removed from the chatbot referring to itself as “MechaHitler” and spewing hate speech, it’s more a bit concerning to imagine Grok handling classified military data. Even after that incident, Grok ranked near the bottom of the list in a Crisis Assessment and Response Evaluator (CARE) test from the team at Rosebud.
