Musk v. Altman OpenAI Trial Testimony

Trials often turn on what a defendant says under cross. This one is turning on what the plaintiff said. Musk took the stand first, repeated 'You can't just steal a charity' as a near-mantra, and walked the jury through what his lawyers framed as a charitable-trust violation: OpenAI was founded as a 501(c)(3) to keep AGI out of any single party's hands, the argument goes, and the for-profit conversion converted donor-funded assets to private gain.

The problem is the same testimony delivered defense-grade material. Musk acknowledged on the record that xAI 'distills OpenAI's models to train its own models' — the kind of admission that rebrands a charitable-trust complaint as competitive warfare in the jury's mind. Kalshi's traders moved in real time: Musk's win probability was around 60% as the trial opened and slid to roughly 34-40% over the following days. Single witnesses rarely move prediction markets that much, and the witness moving them was the plaintiff.

Greg Brockman's testimony delivered the trial's most cinematic moment and, more importantly, its most damaging frame for Musk's case. Brockman told the jury that when Musk demanded full and absolute control of OpenAI in August 2017, he gave a specific reason: he needed roughly $80 billion to build a self-sustaining city on Mars, and OpenAI was a vehicle that could plausibly throw off that kind of capital. When the founders refused, Brockman testified, Musk stormed around the table; 'I thought he was going to hit me.'

That anecdote does something the defense couldn't manufacture: it converts Musk's stated reason for breaking with OpenAI from 'OpenAI is on a path of certain failure' (his February 2018 framing) into 'OpenAI wouldn't give me what I needed for Mars.' The legal claim Musk is bringing is fiduciary in nature — he says he donated $38 million believing it would stay in a nonprofit. But Brockman's 2017 anecdote, paired with Musk's own 'distill' admission, lets Judge Gonzalez Rogers read the entire arc as a thwarted control bid rather than a defense of mission. Even Fortune's Jeremy Kahn, who's covered the case throughout, calls it 'a distraction, generating much more heat than it is shedding light on the bigger concerns about who controls AI.'

The most consequential single document in the trial may not have come from any witness. On April 25 — two days before jury selection — Musk texted Brockman gauging settlement interest. Brockman countered by suggesting both sides drop all claims. Musk's reply, now sought by OpenAI's lawyers as exhibit material via Brockman's live testimony rather than a screenshot: 'By the end of this week, you and Sam will be the most hated men in America. If you insist, so it will be.'

The legal mechanics here matter. Federal Rule of Evidence 408 normally bars settlement communications from being admitted to prove liability — and on Reddit the most-upvoted thread on the trial is largely a debate among self-identified attorneys over whether OpenAI's filing route preserves admissibility. But OpenAI's filing isn't trying to use it for liability; it's trying to use it for motive. If Judge Gonzalez Rogers admits it, the case effectively becomes Musk-the-plaintiff arguing he sued to defend a charity, with a contemporaneous text in evidence saying he sued to make Altman and Brockman the most hated men in America. That kind of motive evidence is exactly the kind a fact-finder weighs against a request for $150 billion in equitable relief.

Mira Murati's video testimony and the November 19, 2023 text chain with Altman — both entered into evidence — open a parallel storyline that the legal outcome doesn't close. Murati, who served 48 hours as interim CEO during Altman's 2023 board ouster, told the court her concern at the time was 'about Sam saying one thing to one person and completely the opposite to another person.' The texts themselves capture Altman pleading with Murati during the firing ('still don't want me?') and Murati relaying the board's position ('They don't want you. They're convinced about their decision.'). Altman is also recorded asking whether Microsoft could acquire OpenAI as an exit ramp.

None of that is, strictly, what Musk is suing about. Musk's complaint is about the 2019-2026 nonprofit-to-PBC conversion. But the texts and Murati's framing function as a permanent, court-stamped record of Altman's 2023 management style — the kind of evidence that lives forever in board reviews, partner conversations, and founder back-channels even if the jury's advisory verdict goes OpenAI's way. The reputational damage is uncorrelated with the legal damage, which is why community sentiment on Reddit is overwhelmingly skeptical of Musk's lawsuit but still treats the trial as a referendum on Altman.

The financial scaffolding around the trial is striking precisely because it doesn't line up. Musk pledged $1 billion to OpenAI and contributed $38 million in actual donations — confirmed on the stand and by Wikipedia's restatement of the founding pledge. He is asking for roughly $150 billion. Brockman's recently disclosed equity stake in OpenAI is reported at near $30 billion. OpenAI itself is privately valued north of $850 billion. Brockman's own journal entries reference 'over $150 billion of OpenAI equity value' tied to the nonprofit — a number Musk's team is now using as the anchor for damages.

The gap between what was given (~$38M) and what is being demanded (~$150B) is the gap a charitable-trust theory has to bridge. Musk's argument is that he is not asking for the return of his donation but for a remedy proportional to the asset class that was misappropriated — i.e., the nonprofit's claim on present-day OpenAI equity. That's a defensible legal theory in the abstract, but it requires the court to find that the 2019 capped-profit structure, blessed by counsel and disclosed at the time, was a breach Musk did not consent to. Two factors complicate that finding: he stayed publicly involved with OpenAI through 2018, and his own February 2018 'certain failure' framing implies he saw it as a failed nonprofit, not a charity to be policed. The damages math is large, but the path to it runs through findings of fact that the plaintiff's own historical conduct keeps undercutting.