Musk Confirms xAI Used OpenAI’s Models to Build Grok—Calls It Standard Industry Practice

Elon Musk admitted under oath Thursday that his AI company, xAI, used distillation techniques on OpenAI’s models to train its Grok chatbot. The revelation came during a California federal court hearing in Musk’s lawsuit against OpenAI, CEO Sam Altman, and Greg Brockman, where he alleges the company abandoned its original nonprofit mission for a for-profit structure.

Distillation—a process where developers query public chatbots or APIs to extract knowledge and train cheaper, nearly-as-capable models—has been a hot-button issue. Industry attention has focused on Chinese firms using the method to undercut U.S. offerings. But many engineers assumed American labs also employ it to keep pace. Musk’s testimony confirms those suspicions. When asked directly whether xAI had distilled OpenAI models, he replied, “Partly,” adding it’s a widespread practice across AI companies.

The admission carries irony. Frontier labs have themselves faced scrutiny for scraping copyrighted data to train their models. Now they’re concerned distillation erodes the competitive edge built on massive compute investments. OpenAI, Anthropic, and Google have reportedly formed a working group through the Frontier Model Forum to detect and block systematic model queries from China.

During his testimony, Musk also ranked today’s top AI labs: Anthropic first, followed by OpenAI, Google, and Chinese open-source models. He described xAI as a much smaller player with only a few hundred employees.

Source: TechCrunch