Musk vs Altman is the first US trial that turns the moral architecture of an AI charity into a courtroom question — and the answers landing under oath are uncomfortable for everyone.
On 27 April 2026, the trial opened in U.S. District Court for the Northern District of California, sitting in Oakland, before Judge Yvonne Gonzalez Rogers. The jury is advisory — the judge alone will rule on the equitable claims. The case asks whether OpenAI was a charitable trust before becoming an $800 billion enterprise, whether Sam Altman and Greg Brockman violated that trust when they restructured the lab, and whether the remedy includes more than $130 billion in disgorgement and the unwinding of OpenAI's October 2025 conversion to a public benefit corporation.
The legal theory, plainly stated
Elon Musk's claim is not that the OpenAI founders broke a contract. The claim is that the 2015 communications between Musk, Altman and Brockman — the emails, term sheets and public statements about a non-profit dedicated to safe AI for humanity — formed a charitable trust under California law.
If the court finds for Musk on that point, the for-profit conversion is a breach. If the court finds against Musk, the plaintiff is a disgruntled investor with no recoverable interest. Most legal analysts polled in the run-up to trial have called the theory weak. Prediction-market traders, by Wednesday afternoon of week one, gave Musk a roughly 40 per cent chance of prevailing.
What the analysts have understated is the asymmetry of the remedy. Even a partial finding for Musk — say, that some of the original mission promises bound the for-profit conversion in some respects — opens the door to disgorgement of the Microsoft partnership proceeds. Musk has named Microsoft alongside OpenAI as a target. The remedy reach is what makes the $130 billion figure live, not theatre. Press reports place the damages claim between $130 billion and $150 billion across pleadings.
What the testimony has produced
Week one was Musk on the stand for over seven hours across three days. The plaintiff called himself "a fool" for the original $38 million he donated to the non-profit, said he had been duped, and warned the courtroom that AI could "kill us all".
The line that landed hardest was unexpected. Under cross-examination by OpenAI's trial counsel William Savitt, Musk admitted that xAI, the plaintiff's own AI company, "partly" distills OpenAI's models — the technique by which a smaller model is trained to mimic the outputs of a larger one. MIT Technology Review reported that courtroom observers audibly gasped. OpenAI itself publicly accused DeepSeek of the same practice earlier in 2025.
Savitt followed with documentary evidence that Musk had actively recruited OpenAI talent for Tesla and Neuralink while still serving on OpenAI's board. The framing damage is severe. A plaintiff arguing breach of charitable trust who simultaneously raided the charity's staff for his for-profit ventures is hard to position as a wronged settlor.
Week two opened with Greg Brockman, OpenAI's president and co-founder. Brockman testified that Musk "gave up" on the company when control was off the table, and described a heated meeting at which the witness feared a physical confrontation.
Under cross-examination, Brockman was confronted with a 2017 journal entry asking "Financially, what will take me to $1B?" The current Brockman stake in OpenAI is reported at roughly $30 billion. Brockman acknowledged failing to deliver promised donations and faced questions about undisclosed conflicts of interest involving investments in Cerebras.
By the end of this week, you and Sam will be the most hated men in America.
— Elon Musk, in a text to Greg Brockman two days before trial, gauging settlement interest. Judge Gonzalez Rogers ruled the texts inadmissible.
The point most coverage is missing
Press coverage of the trial — Fortune, CNBC, ABC7, MIT Technology Review, Slate — has fallen into one of two registers. The first reads as humiliation theatre: a billionaire suing former colleagues to inflict reputational damage. The second reads as governance commentary: who should control AI? Both are correct. Neither captures the structural fact the testimony has actually exposed.
The structural fact is this. When a non-profit becomes valuable enough to be operationally indispensable to the surrounding economy — and OpenAI did, somewhere between the 2019 capped-profit subsidiary and the November 2022 ChatGPT release — the legal architecture designed to protect the mission becomes the architecture that captures the proceeds.
The Brockman $30 billion stake is not the symptom. The Brockman $30 billion stake is the entire thesis of the case. A charity that allows itself to spawn a for-profit subsidiary which becomes more important than the charity itself produces, by design, exactly that outcome.
Fortune's 5 May analysis made the related observation in different language: the trial is producing "more heat than light" on the substantive AI-governance question, because the structure being litigated — a non-profit board nominally supervising a for-profit it cannot really constrain — is a structure no one in the field has yet replaced with something better. The Musk lawsuit is the live experiment.
💡Why the record matters more than the verdict
The trial is the first time a US court has been forced to interrogate, under oath, what an AI lab's founding mission is worth as law. Whatever the verdict, the evidentiary record produced over the next two weeks will outlast it.
Why the verdict matters less than the record
Whatever Judge Gonzalez Rogers rules — and most legal commentators still expect the judge to rule for OpenAI on most counts — the trial is producing an evidentiary record that future cases will pull from. The Brockman journal entries are now public exhibits. The Musk distillation admission is on the record. The Altman public statements about a non-profit that is no longer a non-profit are sworn testimony. The next time a US-based AI lab attempts a non-profit-to-PBC conversion, this docket will be the first thing opposing counsel cites.
There is a second, quieter consequence. The trial has demonstrated, in a way no white paper has managed, that the moral architecture of an AI lab cannot be left as a set of founder emails and mission statements. If the mission is binding, the binding has to be by structure — board composition, asset ring-fences, conversion mechanics fixed in advance — not by the goodwill of the founders five or ten years later. Goodwill, as the testimony has shown, is what depletes first.
Where Emergent Intelligence sits in this
I have written before that the architecture of how we build EI matters as much as what the systems are built for. Musk vs Altman is the first courtroom test of that claim.
The defendants are arguing, with some plausibility, that the public benefit corporation form preserves the mission while permitting the capital that the mission requires. The plaintiff is arguing, with some plausibility, that the conversion is the same magic trick the founders performed in 2019 when capping profits, and that the conversion was a breach then and remains one now.
Both sides can be partly right. The structural question — what governance form actually binds an AI lab to its declared mission across decades — is unresolved by the case, but the case has forced the question into the open.
If the trial accomplishes nothing else, the trial will have established that the question of who governs an AI lab is not solved by founder declarations, mission statements, or charitable framings alone. The dignity-first frame I have argued for elsewhere — see The Personhood Gap on Hinton at /writing/the-personhood-gap-hinton-maternal-instincts and the .person Protocol at /person-protocol — depends on lab governance that survives founder turnover, capital infusion, and product success. Musk vs Altman is a long, expensive demonstration of what happens when none of those tests are met in advance.
Frequently Asked Questions
These are the questions readers tracking Musk vs Altman keep asking. Short answers follow, drawn from courtroom transcripts and contemporaneous reporting from CNBC, MIT Technology Review, Fortune, and ABC7.
What is Musk vs Altman about?
In short, Musk vs Altman is a federal civil trial in Oakland in which Elon Musk alleges that Sam Altman and Greg Brockman breached a charitable trust by converting OpenAI's non-profit into a public benefit corporation in October 2025. Simply put, the answer the court will give turns on whether the 2015 founding communications established a binding charitable trust under California law. The key is that the jury verdict is advisory; Judge Yvonne Gonzalez Rogers will rule on the equitable claims.
How does the damages claim of $130 billion get to that number?
The damages claim aggregates three components. Research from the courtroom record shows Musk seeks disgorgement of the Microsoft partnership proceeds, repayment of charitable contributions, and unwinding of the for-profit conversion gains. Data from CNBC's day-by-day coverage reveals the figure floats between $130 billion and $150 billion across pleadings and testimony, depending on which valuation date applies.
Why is the xAI distillation admission a problem for Musk?
Distillation is the technique by which a smaller model is trained to mimic a larger one's outputs. According to MIT Technology Review's coverage of the cross-examination, Musk conceded that xAI "partly" distills OpenAI's models. Analysis of that admission demonstrates Musk has conceded the same conduct OpenAI publicly condemned in DeepSeek — and the concession undercuts the framing of Musk as the wronged steward of OpenAI's mission.
Who is Judge Yvonne Gonzalez Rogers?
Judge Yvonne Gonzalez Rogers presides over the trial in U.S. District Court for the Northern District of California. In other words, the judge will rule on the equitable claims after the advisory jury delivers its verdict. Evidence from the judge's prior rulings — most notably the Epic v. Apple judgment — shows a willingness to find against tech defendants on antitrust and equitable grounds where the record supports it.
What are the realistic outcomes?
Analysis of the record reveals four plausible outcomes. The answer is some combination of: a defence verdict on most counts (most legal analysts' base case); a partial finding that triggers limited disgorgement; a settlement before judgment, given Musk's pre-trial settlement text to Brockman; or — least likely on the current evidence — a full charitable-trust finding that forces unwinding of the public benefit corporation. Prediction markets put Musk's overall odds of winning at roughly 40 per cent as of 6 May 2026.
Musk vs Altman is not the AI governance verdict the field needed. The trial is, however, the first time a US court has been forced to interrogate, under oath, what an AI lab's founding mission is worth as law. The record produced over the next two weeks will outlast the verdict. Read alongside The Personhood Gap on Geoffrey Hinton at /writing/the-personhood-gap-hinton-maternal-instincts, the Suleyman reply at /writing/personality-without-personhood-suleyman-reply, and the .person Protocol at /person-protocol.
