Musk's Lawsuit Against OpenAI Stumbles in Court After xAI Training Admission

The first week of Elon Musk's trial against OpenAI did not follow his script. Instead of a simple story about a "betrayed mission," the Oakland hearings quickly turned into a review of facts awkward for Musk himself, from old internal documents to the admission that xAI used OpenAI models to train Grok.

What came out in court

The trial opened on April 28, 2026, in federal court in Oakland. Formally, the case has a nine-person jury, but its verdict is advisory: the final decision on the parties' liability and possible remedies rests with Judge Yvonne Gonzalez Rogers. That matters because the case looks less like a classic contest for jurors' sympathy and more like a public audit of OpenAI's entire history, its correspondence and the motives of those involved.

Over three days of testimony, Musk laid out his familiar version of events: he helped launch OpenAI as a nonprofit for the sake of AI safety, and Sam Altman and Greg Brockman then allegedly turned it into a commercial machine. But on cross-examination, OpenAI attorney William Savitt showed documents from 2017–2018 indicating that Musk himself also pushed for a move to a commercial structure, under his own control. That weakened the lawsuit's central claim that Musk had consistently and from the start backed only a nonprofit model.

Awkward admissions

The most notable moment of the week was Musk's admission that xAI trained Grok, at least in part, using outputs from OpenAI models. In court this is not a separate charge, but it is a heavy blow to the plaintiff's public stance. For many months Musk has criticized OpenAI for abandoning its original mission and concentrating power, and it has now emerged that his own company uses the same ecosystem as a source for training.

The irony is not only reputational. Distillation, the extraction of useful patterns from another model's answers, has become one of the most sensitive issues for big AI labs. OpenAI, Anthropic and others are publicly fighting the practice, especially when it involves mass copying of model behavior through APIs. Against that backdrop, Musk's answer of "partly" looked especially awkward: in effect, he confirmed that xAI does what the market is now actively building defenses against.

• The court received documents in which Musk himself discusses OpenAI's commercial pivot
• xAI admitted to partial training on outputs of OpenAI models
• OpenAI is trying to show that Musk filed some of his claims too late
• Before the trial began, the judge removed the fraud charges from the case
• Both sides were barred from turning the hearings into a debate about the end of humanity because of AI

"This is not a trial about the risks of artificial intelligence to

humanity," the judge essentially reminded the parties.

The judge repeatedly steered participants back to the narrow subject of the dispute: whether OpenAI breached its contractual and fiduciary obligations when it restructured the company. So far, attempts by Musk and his team to broaden the discussion to the philosophy of AI safety have not worked. The court even noted an obvious contradiction: if Musk truly considers such systems an existential threat, why does his own company operate in the same sector and by the same growth rules?

What is at stake

Musk is seeking not only monetary compensation, which various reports put at more than $130 billion, but also structural changes: returning OpenAI to a nonprofit format and removing Altman and Brockman from leadership roles. OpenAI responds that creating a commercial arm in 2019 was a necessary condition for buying computing capacity and hiring researchers, without which the company could not have competed with Google and other players. OpenAI is now structured as a public benefit corporation in which a nonprofit entity and investors, including Microsoft, hold stakes.

The trial still has several weeks and heavier witnesses ahead. Testimony is expected from Sam Altman, Greg Brockman, Microsoft CEO Satya Nadella and early OpenAI engineers. Their cross-examinations should clarify whether OpenAI had a real contractual ban on commercialization or whether this was more a clash of expectations between co-founders. For now, week one looks like this: instead of a prosecutorial march against OpenAI, the court has begun exposing Musk's own internal contradictions.

What this means

For the AI market, this is no longer just a dispute between former partners. The Oakland court is testing whether an AI lab can be legally held to its original public mission when giant investments, compute and the race for leadership are at stake. And the first week showed that the moral argument works worse if your own company plays by the same rules.