Elon Musk Appears to Acknowledge xAI’s Use of OpenAI’s Models for Its Training

During his testimony in federal court on Thursday, Elon Musk suggested that his AI lab might have utilized OpenAI’s models to train xAI’s own systems. He brought up this point while on the witness stand, responding to cross-examination questions from an attorney representing OpenAI amid his ongoing legal dispute with the organization that created ChatGPT.

Here’s the conversation captured by WIRED:

OpenAI Lawyer William Savitt: Are you familiar with the concept of distillation?

Musk: It refers to using one AI model to train another.

Savitt: Has xAI employed this technique with OpenAI?

Musk: Generally, most AI companies do that.

Savitt: So that confirms it as a yes.

Musk: Partially.

Distillation is a process in which a smaller AI model is trained to emulate the behavior of a larger, more advanced model, allowing for more efficient and rapid execution while maintaining a significant level of performance.

Next, OpenAI’s lawyer, William Savitt, inquired whether OpenAI’s technology had contributed in any manner to the development of xAI.

Savitt: Has OpenAI’s technology been utilized in any way for xAI’s development?

Musk: It’s common practice to use other AIs for validating your own.

Neither OpenAI nor xAI offered immediate comments to WIRED’s inquiry.

OpenAI has been actively working to deter competitors from distilling its AI models, particularly targeting the Chinese AI lab DeepSeek. In a memo addressed to a House committee in February 2026, OpenAI stated that it has “taken steps to protect and harden our models against distillation.” The memo emphasized a commitment to maintaining a competitive landscape where “China cannot advance autocratic AI by appropriating and repackaging American innovation.”

The Trump administration has also initiated measures to prevent Chinese companies from distilling American AI models. Michael Kratsios, director of the White House’s office of science and technology policy, mentioned in an April 2026 memo that information would be shared with U.S. AI firms regarding foreign distillation efforts. Kratsios reiterated in a post on X that the “U.S. government is dedicated to the free and fair evolution of AI technologies within a competitive ecosystem.”

American AI labs have collaborated with each other’s models in various ways to evaluate progress and ensure safety. However, in the current competitive environment, some companies have entirely severed ties with rival laboratories. In August 2025, Anthropic restricted OpenAI’s access to its Claude coding models, claiming a violation of terms of service. Recently, Anthropic has denied xAI access to its coding models as well.

Throughout his extensive cross-examination of Musk, Savitt has probed Musk regarding his attempts to gain control of OpenAI and his subsequent ambitions to outperform the ChatGPT creator. On Wednesday, Savitt presented emails and messages from 2017 to substantiate inquiries about whether Musk pressured OpenAI by withholding funding and recruiting significant researchers.