The Oakland verdict in Musk v. Altman came in this morning at the federal courthouse on Clay Street. A nine-person advisory jury, sitting before Judge Yvonne Gonzalez Rogers in the Northern District of California, took less than two hours to find unanimously that Elon Musk's claims against OpenAI, Sam Altman, Greg Brockman, and Microsoft were filed outside the three-year statute of limitations. The judge accepted the verdict as the court's own. The case is over at the trial level. Musk posted on X within minutes of the verdict that the decision was a "calendar technicality" and that he would appeal to the Ninth Circuit.
The framing competition began before the lawyers had cleared the courtroom. Musk on X: "the judge & jury never actually ruled on the merits of the case, just on a calendar technicality." OpenAI's lead counsel William Savitt to reporters on the courthouse steps: "not a technical decision, it's a substantive one." Both statements are true at different levels of the operative facts, and the level at which Savitt's is true is the one this piece is about. The procedural dismissal Musk is calling a technicality is, in the structure of California charitable-trust law, the substantive determination that founders and donors cannot wait six years to bring private claims against a charity for mission drift. The court is not saying the charity wasn't violated. The court is saying that the question of whether it was violated cannot be answered by Musk, in this forum, on this timing.
What the verdict actually held
The jury's finding was narrow and clean: Musk's two surviving claims — breach of charitable trust against OpenAI, Altman, and Brockman, and a parallel aiding-and-abetting theory against Microsoft — fell outside the three-year window the California statute affords. Judge Gonzalez Rogers, in confirming the verdict as the court's own, made no finding on whether the founding-mission claim had merit. The breach-of-charitable-trust theory the trial was nominally about was never reached.
The deliberation time matters as evidence about how clean the procedural defense looked from inside the jury room. Two hours is shorter than the lunch break would have run on a contested merits question. Three weeks of testimony, including roughly four hours from Altman in which Musk's lead trial counsel Steven Molo opened cross-examination with the question "Are you completely trustworthy?", plus live trial testimony from Brockman, Sutskever, Nadella, Microsoft CTO Kevin Scott, and former OpenAI board member Shivon Zilis, all converged on a procedural question that the jury found dispositive in less time than it would take to drive from the courthouse to Cupertino.
The 26-to-zero attrition
The trial verdict is the last station in a multi-stop pretrial attrition of Musk's case. Musk first filed in San Francisco Superior Court in February 2024, withdrew that complaint, and refiled in federal court in the Northern District of California in August 2024. The federal complaint and its amended versions grew to twenty-six claims at their peak. The version that went to the jury was the third amended complaint, and the surviving claims at trial were three of the original twenty-six. The path, summarized:
| Claim category | Dismissal stage | Mechanism | What it would have required |
|---|---|---|---|
| Breach of contract | Pretrial (March 2025) | Motion to dismiss granted | A written founding agreement Musk could plead; the court found none |
| RICO racketeering | Pretrial | Motion to dismiss granted | A pattern of predicate acts the court found unsupported |
| Antitrust | Pretrial | Motion to dismiss granted | A defined market and an exclusionary act; the court found neither pled |
| False advertising | Pretrial (May 2025) | Motion to dismiss granted | A specific commercial statement and reliance |
| Breach of fiduciary duty | Pretrial (May 2025) | Motion to dismiss granted | A fiduciary relationship the court declined to find |
| Tortious interference (vs Microsoft) | Summary judgment (January 2026) | Granted to Microsoft | An interfered-with contract; court found no actionable contract |
| Unjust enrichment (vs Microsoft) | Summary judgment | Granted to Microsoft | A direct benefit conferred on the defendant by the plaintiff |
| Breach of charitable trust | Jury verdict (May 18, 2026) | Statute of limitations | A claim filed within three years of knowledge of breach |
| Unjust enrichment (vs OpenAI/Altman/Brockman) | Jury verdict | Statute of limitations | Same three-year window |
| Aiding and abetting (vs Microsoft) | Jury verdict | Statute of limitations | Same three-year window |
The table is the procedural shape of the case. Each row is a step at which Musk's theory needed to clear a different doctrinal bar. The first five rows fell at the pleading stage on substantive doctrinal grounds. The middle two fell at summary judgment, again on substantive grounds. The last three fell at the jury stage on procedural grounds. The cumulative reading is not that Musk had one strong claim that was defeated by a clock. The cumulative reading is that the claims with substantive teeth never made it out of pretrial, and the claims that did make it out of pretrial fell on the doctrine that gave them their teeth to begin with: charitable-trust law has a statute of limitations because charitable-trust law treats donor recourse as time-bounded.
The standing problem nobody discussed at trial
The structural fact the courtroom argument largely set aside is that California has no donor-standing statute. The state's regulatory framework for public benefit corporations, anchored in California Corporations Code §5142 and the Government Code's Attorney General supervision provisions at §12580 et seq., enumerates a narrow list of plaintiffs who may bring an action to remedy a breach of charitable trust: the corporation itself, a member, an officer or director, "a person with a reversionary, contractual, or property interest" in the corporation's assets, and the Attorney General. Donors as a class do not appear in the list. The California Attorney General's Guide for Charities makes the same point: donor enforcement of charitable-purpose terms is, as a default rule, not actionable in California state-law tradition.
Musk's pitch to the court was that he had standing through two doors at once. The first was contractual: that the founding agreement he says existed in 2015 gave him a contractual interest in the charity's purpose. That door closed in March 2025 when the court dismissed the breach-of-contract claim for failure to plead a sufficient written contract. The second door was the charitable-trust claim itself, on the theory that he was bringing it not as a contractual creditor but as a person acting in the interest of the charitable beneficiary class. That door is the one the jury closed today on timing grounds, but the door was structurally narrow before the jury arrived. California charitable-trust enforcement is the Attorney General's province, by design. The two-hour deliberation is the artifact of that design.
The framing question Musk is now publicly relitigating — was the dismissal "technical" — looks different inside that frame. The procedural dismissal is the substantive design of California charitable-trust law operating as designed. The dismissal is technical only in the sense that all law that does its job through doctrine is technical. The dismissal is substantive in the sense that doctrine is how the law decides who is allowed to enforce what.
The Sutskever email
The piece of evidence the trial did not need to resolve but that the public record now contains is the email exchange OpenAI itself published on its OpenAI and Elon Musk page and that was entered as an exhibit at trial. On January 2, 2016, OpenAI co-founder and then-chief-scientist Ilya Sutskever wrote to Musk: "As we get closer to building AI, it will make sense to start being less open. The Open in openAI means that everyone should benefit from the fruits of AI after its built, but it's totally OK to not share the science..." Musk's reply: "Yup."
The exchange is structurally consequential because the claim Musk asked the jury to ratify was that he had been induced to seed OpenAI on the explicit promise that any AGI developed there would be open and shared with the world, and that the conversion away from that openness constituted the breach. The Sutskever email is contemporaneous evidence that the conversion-away-from-openness was a discussion Musk was part of and assented to in writing, eight years before he sued. A merits-stage finding on whether the founding-mission promise was breached would have had to reckon with that email. A statute-of-limitations dismissal does not need to.
This is the part of the verdict that maps to Savitt's "substantive" framing in the cleanest possible way. The procedural defense the jury accepted is not a workaround the defense found at the last minute. It is the doctrinal expression of a structural fact: by early 2016, the question Musk later sued over was already in his inbox and already had his Yup on it.
What the Attorney General has not yet done
The institutional actor the verdict implicitly defers to is the California Attorney General. The October 28, 2025 OpenAI restructuring — under which the nonprofit became the OpenAI Foundation, the for-profit became OpenAI Group PBC, and the Foundation took an equity stake reported at roughly $130 billion in the new entity — went through standard charitable-trust review under the AG's office before completing. The AG signed off. That signoff is the substantive charitable-trust-enforcement decision in the operative regulatory regime; the trial verdict today is the substantive donor-standing decision in the doctrinal regime that frames who else can challenge it.
The follow-up question the verdict leaves open is whether the AG's process is adequate to the stakes. The dollar exposure the case was tried on was up to $150 billion of disgorgement to the nonprofit. The AG-supervised restructuring placed the nonprofit at the head of a $130 billion equity ladder it controls but does not solely capture. Whether those two numbers describe the same charitable interest, and whether the AG's signoff bound the institutional question for purposes other than this private case, are questions that will live in academic and regulatory writing for several years. The trial verdict does not answer them. The trial verdict makes clear that founders and donors cannot answer them via private litigation under existing California law.
This is the part of the story most worth carrying forward. The verdict's substantive content is not Altman won and Musk lost. The verdict's substantive content is charitable-trust enforcement against major AI labs in California runs through the Attorney General, and the Attorney General has already adjudicated the structural question the trial was nominally about.
The appeal that probably won't change the outcome
Musk's announced appeal goes to the Ninth Circuit. The standard the appellate court will apply to a statute-of-limitations dismissal is de novo on the legal question of whether the limitations period was correctly applied, and clearly erroneous on the underlying factual determinations the trial court adopted from the jury. The legal question is the clock-start determination: when, as a matter of law, did Musk know or should have known of the conduct he is now suing over.
The defensible clock-start anchors the trial record contains are three: Musk's departure from OpenAI in February 2018; his last monetary contribution in 2020; and the public 2019 announcement of the capped-profit conversion that he later characterized as the breach. Any of those anchors, applied to a 2024 filing, places the claim outside the three-year window. The Ninth Circuit would need to find that none of those anchors was knowable to Musk, or that the breach was a continuing one that the 2025 PBC recapitalization tolled forward into the limitations window. The continuing-breach theory is the most plausible path the appeal has, and it is not a path the trial court rejected on technical grounds. It is a path the trial court rejected because, under California charitable-trust precedent, mission drift that the plaintiff was aware of in 2018 does not start a fresh clock at every subsequent corporate restructuring.
The appeal will be heard. The appeal is, on the available record, unlikely to produce a result different from the one Judge Gonzalez Rogers entered today.
Coda
The verdict that came in this morning was not a procedural footnote on a substantive question the court declined to reach. The verdict was the substantive determination that California charitable-trust enforcement runs through the Attorney General and through the three-year window the statute prescribes for the small class of plaintiffs who can act in its absence. Musk filed outside that window, did not pursue the Attorney General path, and lost at the trial stage on the doctrine that gave shape to his only available theory. The calendar technicality framing is a courtroom-press-conference framing. The structural framing is that the doctrine did exactly what doctrine does.
The question the verdict deliberately did not answer is the one that the next several years of AI-governance writing will be about. Whether the AG's review of the October 2025 recapitalization was adequate to the structural stakes of a $130 billion charitable equity position held in an AI lab with eight-figure daily inference revenue is a question for the regulatory regime, not for private litigation. The verdict makes that boundary explicit. Reading the verdict as a personal defeat for Musk is reading the easier story. The harder story is the one about which institution gets to decide what the charity was for, and that institution is not, by design, the founder.
Top comments (0)