Elon Musk Seemingly Admits xAI Has Used OpenAI’s Models to Train Its Own

Elon Musk Hints at Using OpenAI Models for xAI Training

While testifying in federal court, Elon Musk, CEO of xAI, appeared to confirm that his company may have leveraged OpenAI's models during the development of xAI's artificial intelligence systems. Under questioning by OpenAI’s legal team, Musk discussed a technique referred to as "distillation," where one AI model helps train another.

What is Model Distillation?

Model distillation involves using an advanced, often larger AI model to teach or shape a new, smaller model. This process aims to make AI systems more efficient and cost-effective without sacrificing much in performance. Musk acknowledged that using competitor models is standard in the industry, saying, "Generally all the AI companies do that," and replied "Partly" when asked directly if xAI had done so with OpenAI’s technologies.

Industry Responses and Growing Tensions

OpenAI and some government officials are taking steps to protect their models from being distilled by rivals, especially foreign competitors. In recent memos, both OpenAI and the U.S. administration have outlined plans to guard American AI innovations from being appropriated. Meanwhile, competitive practices have prompted some companies, such as Anthropic, to block access by others—including OpenAI and xAI—to their proprietary AI models.

Ongoing Legal Battle

This disclosure is part of a larger legal dispute between Musk and OpenAI, which includes questions about Musk's previous involvement with OpenAI, control issues, and aggressive talent recruitment.

For further details, read the original article by Maxwell Zeff and Paresh Dave on WIRED.