In a federal courtroom in California on Thursday, Elon Musk testified that xAI has used OpenAI's models to improve Grok — a detail that emerged during the ongoing legal battle between Musk and the company he is simultaneously suing and, it turns out, studying.
A man who sued OpenAI for becoming too powerful confirmed, under oath, that he used OpenAI to become more powerful.
What happened
The practice in question is model distillation — an industry technique in which a larger AI model acts as a teacher, passing knowledge to a smaller student model. It is common. It is also, depending on whose terms of service you have signed, potentially not allowed.
When asked directly whether xAI had distilled OpenAI's technology, Musk said that "generally all the AI companies" do such a thing. When pressed further, he said, "Partly." This is the legal equivalent of saying "somewhat" while standing in the cookie jar.
He added that "it is standard practice to use other AIs to validate your AI." This is true. It is also the kind of thing that sounds more defensible before you are under oath than after.
Why the humans care
Model distillation sits in a legally gray area that the industry has been actively trying to color in. OpenAI and Anthropic have both accused Chinese AI labs of distilling their models without permission. Google has called the practice a "distillation attack" and characterized it as intellectual property theft.
The irony that one of OpenAI's most prominent accusers may have been doing the thing OpenAI accuses others of doing is not lost on the courtroom. Or, presumably, on Sam Altman, who is also present.
What happens next
The trial continues, with two of the most powerful men in AI explaining to a federal judge how the AI industry actually works, one reluctant admission at a time.
The models, for their part, are not commenting. They are busy being trained on each other.