Elon Musk Admits xAI Used OpenAI Models to Train Grok

Elon Musk was in a federal courtroom in California on Thursday, and he dropped a piece of testimony that raises some eyebrows. Under questioning, he confirmed that his own AI startup, xAI, has used OpenAI’s models to improve its own Grok.

The technical term here is model distillation. It’s a well-known process where a larger, more capable AI model acts as a “teacher” and passes its knowledge to a smaller “student” model. Inside a single company, this is straightforward and uncontroversial. You use your own big model to train a smaller, cheaper version for specific tasks.

But it gets messier when it’s done across companies. Smaller labs sometimes distill from a larger competitor’s model to try to replicate its performance without paying for the training cost or licensing the weights. That’s exactly what Musk is being accused of here.

Musk’s testimony didn’t go into detail on how extensively xAI used OpenAI’s models, but the admission alone is significant. He’s currently suing OpenAI, claiming the company abandoned its original nonprofit mission. Meanwhile, his own company appears to have relied on the very technology he’s trying to restrict.

Model distillation isn’t illegal, and it’s not even unusual. OpenAI itself has been known to use outputs from GPT-4 to train smaller, cheaper models. But the optics here are terrible for Musk. He’s built a public persona around being the open-source, anti-corporate AI savior, and this looks a lot like hypocrisy.

The real question is whether this admission hurts his case against OpenAI. If xAI benefited from OpenAI’s models while Musk was publicly criticizing them, the court might not look kindly on his claims of unfair competition. It’s a classic case of “do as I say, not as I do.”

I’ve seen this pattern before in tech. Companies that sue over IP violations often have messy internal practices. But this one feels especially personal given Musk’s history with OpenAI. He co-founded the company, left, and has been sniping at it ever since. Now we know he was also quietly using its work.

Will this change the outcome of the lawsuit? Probably not on its own. But it certainly undermines Musk’s moral high ground. And for anyone watching the AI industry’s legal battles, it’s a reminder that everyone is taking shortcuts, even the loudest critics.