Elon Musk sues OpenAI.
In the first week, the biggest revelation is that xAI distilled ChatGPT.

Elon Musk donated $38 million to help establish OpenAI, which is now valued at $800B. What did he say during the first week of the century-long trial where he sued the “deceptive non-profit organization”?
(Background summary: Sam Altman pledged to invest $600 billion to expand computing power to “accelerate OpenAI’s IPO by the end of the year”; CFO worries about cash running out in five years)
(Additional background: xAI secretly launched Grok 4.3: directly generating Word, PowerPoint, Excel files; Microsoft’s moat is being shattered)

Table of Contents

Toggle

• From passionate support to “certainty of being exploited”: three stages
• OpenAI lawyers’ counterattack: email leaks and open-source conflicts
• The biggest bombshell: xAI admits to distilling OpenAI

He took the stand and said he is a fool. Musk accused Sam Altman and Greg Brockman of plundering the non-profit organization and blocking OpenAI’s restructuring in the federal court in Oakland, California.

From passionate support to “certainty of being exploited”: three stages

Reporter Michelle Kim from MIT Technology’s courtroom coverage describes that Musk divided his relationship with OpenAI into three phases during the trial.

The first phase is “passionate support”: In 2015, he co-founded OpenAI with Altman and Brockman, aiming to create a non-profit AI lab to counterbalance Google. Musk quoted Google co-founder Page’s words as motivation.

He asked Page what to do if AI tries to wipe out humanity, and the response was “just want AI to survive.” Musk said, because of this, he believed a non-profit organization focused on safety was necessary, not profit.

The second phase is “loss of confidence”: He began to doubt whether what Altman told him was true.

The third phase is “certainty of being exploited”: The turning point occurred at the end of 2022. Court records show Musk learned that Microsoft would invest $10 billion into OpenAI, and he messaged Altman: “What the hell is going on? This is deception with a different coat.” From a non-profit experiment focused on safety, it turned into a commercial entity with billions from Microsoft. In Musk’s view, that line had been crossed.

He calculated in court: “I gave them $38 million almost for free, and they built a company now worth $800B.”

Musk’s demand is to remove Altman and Brockman from their positions and revoke OpenAI’s restructuring into a “non-profit + profit subsidiary” model. If he wins, it could directly disrupt OpenAI’s plan for an IPO valued at around $1 trillion.

OpenAI lawyers’ counterattack: email leaks and open-source conflicts

OpenAI’s attorney William Savitt presented two emails from Musk during the trial.

The first, written in 2017, Musk informed Tesla Vice President about poaching OpenAI co-founder Andrej Karpathy, stating: “People at OpenAI want to kill me. But it must be done.”

The second email was sent to Neuralink co-founder, indicating that they could directly poach from OpenAI. Musk’s response was:

“This is a free world, I can’t restrict them from recruiting people from other companies.”

Savitt also pointed out that one of Musk’s core arguments in the lawsuit is that OpenAI has deviated from its non-profit mission and turned into a closed-source commercial company. But xAI itself is also a closed-source profit enterprise. Moreover, xAI sued Colorado’s AI anti-discrimination law in April 2026, which starkly contrasts with Musk’s high-profile claims about AI safety during the trial.

The biggest bombshell: xAI admits to distilling OpenAI

The most damaging moment of the first week was when, under Savitt’s questioning, Musk admitted that xAI “partially” distilled OpenAI’s models.

Distillation, simply put, involves training a smaller AI model to mimic the behavior of a larger model, not by copying code directly, but by learning from the output results of the larger model.

Musk’s explanation was: “Using other AIs to verify your AI is standard industry practice.” This makes some technical sense, but in the context of his accusations that OpenAI profited improperly, the credibility is greatly diminished.

Next week, UC Berkeley computer scientist Stuart Russell will testify on AI safety issues, and Brockman will also appear for cross-examination.