Elon Musk said under oath on Thursday that xAI partly used OpenAI models while developing Grok, turning a widely discussed industry practice into an unusually direct courtroom admission.

What was said

During cross-examination in Musk's lawsuit against OpenAI, CNBC reported that OpenAI lawyer William Savitt asked whether xAI had used OpenAI's models to help develop or test its own systems. Musk replied that this was standard practice in the field. TechCrunch and WIRED both reported a more specific exchange on distillation: after Savitt asked whether xAI had done that with OpenAI, Musk answered, "Partly."

The conservative takeaway is narrower than the headline suggests. Musk did not publicly identify which OpenAI models were involved, how often they were used, or what share of Grok's training relied on them. CNBC's live coverage also framed the answer as part of a broader argument over how AI companies validate and benchmark against rival systems.

Why it matters

The testimony lands as frontier labs are getting more aggressive about blocking model-copying and distillation. OpenAI has described distillation as a risk to its competitive moat, and Anthropic has also restricted rival access to some of its models. Musk's answer offers a rare on-record example suggesting that cross-lab borrowing is not limited to smaller outsiders or overseas competitors.

That does not establish wrongdoing on its own. But it does make the industry's public stance on model copying harder to separate from its private development habits.