Elon Musk Admits xAI Used OpenAI’s Models to Train Grok — Right Before Suing Them for the Same Thing

Elon Musk was on the stand in a California federal courtroom on Thursday, and he dropped a quiet bomb: his own AI startup, xAI, used OpenAI’s models to improve Grok.

Let that sink in for a second. The same guy who’s been publicly roasting OpenAI for years, who’s currently suing them for allegedly betraying their nonprofit mission, just admitted under oath that his company did exactly the kind of thing he’s mad about.

The technical term here is “model distillation.” It’s a standard practice in the industry — one large model acts as a teacher, passing knowledge to a smaller student model. Companies do it internally all the time. But it’s also the go-to shortcut for smaller labs trying to make their model behave like a bigger, better one without doing all the expensive training from scratch.

When asked if he knew what model distillation was, Musk said yes. Then he confirmed xAI had used it with OpenAI’s models. That’s a pretty big deal, because the whole premise of Musk’s lawsuit against OpenAI is that they’ve been too cozy with Microsoft and have abandoned their original open-source promises. Now he’s admitting his own team took a sip from the same well.

Look, I’m not saying distillation is inherently bad. Every major lab does some version of it. But when you’re the guy leading the charge against “closed” AI and then quietly using your competitor’s work to bootstrap your own model, it looks less like principled opposition and more like sour grapes.

What’s even more interesting is what this says about the state of AI competition right now. The top models are so expensive to train that even Musk — with his billions — chose the shortcut. If xAI couldn’t afford to build Grok from scratch without borrowing from OpenAI, what chance do smaller players have?

I’ll be watching to see how this plays out in court. Because if Musk’s own testimony doesn’t undermine his case, I don’t know what will.