The Musk v. Altman trial was supposed to be a fight about a charity. By Tuesday afternoon in Oakland, Elon Musk had turned the witness stand into the most public AI safety lecture of the year.

The world's richest man, the man currently asking a federal jury to take $134 billion away from OpenAI and remove Sam Altman from the company he co-founded, used his first day of testimony to tell jurors that artificial intelligence will be "smarter than any human" as soon as next year. He compared it to raising a "very smart child" who eventually outgrows your control. He recounted being called a "speciesist" by Google co-founder Larry Page for prioritizing human survival over machine intelligence. He explained, in his own words, why he and Altman started OpenAI as a nonprofit in 2015: to keep Page from running the most important technology in human history alone.

Whatever the jury makes of the unjust enrichment claim, the cross-examination, or the charitable trust theory, this is now the most quotable courtroom on artificial intelligence ever convened.

OpenAI's lead attorney, William Savitt, opened the morning trying to flip Musk's narrative. Musk, Savitt argued, never cared whether OpenAI stayed a nonprofit. He cared whether he was running it. When the founders refused to fold the company into Tesla in late 2017, Musk left. The case, Savitt said, is "rich man's regret." Musk wanted to control the most important AI lab on Earth, did not get his way, and is now trying to use the courts to undo someone else's success.

Steven Molo, Musk's lawyer, opened with an entirely different story. Altman and Greg Brockman, with Microsoft's help, "stole a charity." That charity, in Molo's framing, was the only entity in 2015 trying to do safe AI research without commercial pressure. The for-profit conversion did not just enrich the founders. It looted a public-interest trust. If the jury allows it, Molo warned, the verdict "will give precedent to looting every charity in America."

Musk took the stand mid-afternoon. He looked, by every reporter's account, funereal. Black suit, black tie, dark mood. He was the trial's first witness. Brockman, sitting at the defense table 20 feet away, calmly took notes on a small reporter's pad while the world's most internet-poisoned billionaire described, under oath, the man Brockman has worked beside for a decade as a thief.

The substance of Musk's direct testimony was a tour through OpenAI's founding mythology. Page wanted scale and speed. Musk wanted alignment and survival. Larry called Elon a "speciesist" for caring about humans more than digital minds. Musk decided he could not let one company, run by one philosophy, build the technology that would replace human labor. He pitched Altman on a counterweight. They built OpenAI together. The nonprofit structure, Musk said, was the entire point.

Then, on the same stand, Musk pivoted into prediction. AI smarter than the smartest human alive: as soon as 2027. Maybe 2028 if scaling slows. The honesty-and-integrity-and-being-good values you instill in your child, you cannot install in superintelligent software. You can train. You cannot control. The implication, never stated, is that nine years after he and Altman co-founded the company specifically to prevent Larry Page from owning the technology that would outthink humanity, the technology has been handed to Altman, Brockman, Microsoft, and a $500 billion private valuation.

The legal stakes are not theoretical. If the advisory jury recommends, and Judge Yvonne Gonzalez Rogers accepts, a finding for Musk on either remaining count, the remedies phase that begins May 18 could include any combination of $134 billion redirected to OpenAI's nonprofit foundation, removal of Altman and Brockman from their executive roles, and forced reversion of OpenAI to a nonprofit structure. The third remedy would unwind the for-profit conversion that made every prior investor whole. It would, in plain language, kill the IPO.

Kalshi and Polymarket are pricing Musk's win probability at 45%. That is not a long-shot. That is closer to a coin flip than the bull case for OpenAI's IPO needs.

Tuesday's testimony was day one of three weeks of liability hearings. Musk continues on direct Wednesday morning. Cross-examination starts after that. Brockman and Altman are both on the witness list. Satya Nadella is on the witness list. Mira Murati, Ilya Sutskever and Shivon Zilis are all expected to appear. The Brockman 2017 diary entry calling the nonprofit commitment "a lie" is the document around which the entire case turns.

What is being established right now, with every news cycle, is a record. The witness list reads like the founding deposition of modern AI. The trial transcript will be cited in every nonprofit-to-for-profit AI conversion that follows it. Anthropic is on track for an October IPO whose entire structure depends on the legal viability of for-profit conversions of AI safety nonprofits. The Musk v. Altman verdict is direct precedent for whether that conversion holds.

OpenAI knows this. Microsoft knows this. The amended Microsoft deal that got announced Monday morning, the same morning the jury was seated, was specifically structured to give Microsoft an exit ramp from the AGI clause. That was not a coincidence. The entire commercial architecture of OpenAI is being re-papered in real time around the legal risk of this trial.

For the AI Post readership, the part to watch is not the daily soap opera. The part to watch is the witness rotation. Brockman's cross-examination will be the moment the diary entry comes into evidence. Altman's testimony will be the moment Musk's lawyers ask him whether the for-profit conversion was about funding research or capturing equity. Nadella's appearance will be the moment Microsoft gets to explain, under oath, why it agreed to a contract with a charity and ended up with exclusive rights to commercialize the resulting technology.

If you only follow one line of testimony for the rest of the year, follow this one. Tuesday afternoon was the warmup.

Musk testifying that AI will be smarter than the smartest human within a year, while sitting on a witness stand in a charitable-trust case against the company building it, is a cleaner summary of where 2026 actually sits than any analyst report you will read this week. The most public AI safety prediction of the decade just got entered into the federal record. By the same person who founded the company he is suing, in support of taking it back from the people he founded it with.