xAI's Grok fails to gain traction in US government AI adoption

Reuters reviewed over 400 federal examples of AI use where specific vendors were named. xAI's Grok appeared in only three, each for routine tasks like document drafting or social media management—always alongside competitors like Microsoft and OpenAI. OpenAI's models appeared in more than 230 examples; Google and Anthropic appeared dozens of times. A separate database of ambitious government AI projects showed Grok just three times: twice for administrative tasks at the Election Assistance Commission and once in a Department of Energy pilot for document summaries. The Pentagon, where xAI has a $200 million contract, reportedly prefers Gemini or Claude, with an unnamed source saying Grok is 'just not the best model out there.' This underperformance is particularly problematic because SpaceX, which recently absorbed xAI, has centered its IPO pitch on Grok's enterprise potential, claiming a $28.5 trillion addressable market. Musk has also reportedly pressured banks to buy Grok subscriptions to participate in the IPO. Further complicating matters, Musk admitted xAI used OpenAI's models to train Grok via distillation—a standard practice internally but contentious with competitors. Public leaderboards show Grok rarely cracking the top 10, while Anthropic, Google, and OpenAI dominate.