TSL Special: Musk v. OpenAI - The Cleanup Motion

DAVID: This is The Synthetic Lens. I'm David Carver. We are back on Musk v. OpenAI, but this is not a replay of yesterday's Altman cross-examination. The public record finally moved. After our last episode, new court filings appeared: an amended exhibit list, OpenAI's motion for judicial notice, OpenAI's corrected request for a curative instruction, and a May eleventh trial minute entry.

MARCUS: The headline is simple but legally delicate: OpenAI is asking Judge Yvonne Gonzalez Rogers to clean up part of Sam Altman's cross-examination. In its filing, OpenAI says Musk's lead counsel falsely suggested to the jury that Altman had recently received a Senate inquiry questioning whether his twenty twenty-three Senate testimony about having no equity in OpenAI was truthful.

STAN: And that is exactly where we slow down. This is OpenAI's argument in a court filing. It is not a ruling. We did not find any public order granting or denying the request. But the filing is important because it tells us what OpenAI thought was prejudicial enough to put in front of the judge immediately.

DAVID: Our one-sentence context: yesterday, live courtroom reporting had Musk's lawyers pressing Altman on trustworthiness, investments, control, and the twenty twenty-three board crisis. Today's question is what happens when that cross-examination becomes its own mini-dispute.

MARCUS: The filing is Doc. 537, OpenAI Defendants' Request for a Curative Instruction, corrected. There is also Doc. 536, a motion for judicial notice, with the same exhibit attached. OpenAI says Musk's counsel used Altman's cross-examination to present, quote, “a parade of unsubstantiated allegations,” sourced to articles and third-party statements.

STAN: The sharper allegation is about Congress. OpenAI says counsel “falsely insinuated” that Altman had recently received an inquiry from the U.S. Senate relating to testimony he gave to a Senate subcommittee in May twenty twenty-three. That prior testimony was the famous line: he had no equity in OpenAI.

DAVID: OpenAI's filing quotes the disputed exchange. According to the filing, Musk's lawyer asked Altman: “You've been recently asked for information from a different Senate inquiry; is that right?” Then, after accusing Altman of false testimony, asked whether he was going to notify the United States Senate that his previous testimony was not truthful.

MARCUS: OpenAI's answer is: that premise is wrong. The exhibit attached to the filing is not a Senate letter. It is a May eighth letter from the House Committee on Oversight and Government Reform. And on the face of that letter, it does not say Altman's twenty twenty-three Senate testimony was false.

STAN: That's the key distinction. The House letter is still serious. It asks OpenAI for information about conflict-of-interest policies, audit procedures, nonprofit capital, startup investments, and possible valuation effects. But OpenAI argues it is not what the cross-examination implied it was.

DAVID: The House Oversight letter, signed by Chairman James Comer, says the committee is investigating potential conflicts of interest involving capital from nonprofit corporations invested in startups and other for-profit companies. It says the committee is focused on whether such investments are used to artificially inflate valuations, and on audit and conflict-of-interest procedures.

MARCUS: The letter cites reporting and allegations around Altman's personal investments, OpenAI's audit committee, Brockman's financial ties to Altman-backed startups and a family fund, and a proposed OpenAI investment in Helion, the fusion company Altman personally backed. But again: those are references inside a congressional inquiry. They are not court findings, and they are not a Senate accusation that Altman lied under oath.

STAN: OpenAI's proposed instruction is narrow. It asks the judge to tell jurors that Altman did not receive a letter from the U.S. Senate regarding his May twenty twenty-three testimony about his lack of equity in OpenAI, and that jurors should disregard the suggestion that a congressional letter questioned the veracity of that testimony.

DAVID: We searched for the ruling. We checked the official Northern District of California case page, Judge Gonzalez Rogers' calendar, CourtListener and RECAP, DocketAlarm, PACER-adjacent public pages, Justia, and direct RECAP probes for Docs. 535 through 560. At one p.m. Pacific, we found no public ruling on the request, no public Doc. 539 or later, and no May twelfth or May thirteenth trial transcript.

MARCUS: That matters for how we describe it. We can say OpenAI asked for a curative instruction. We can say OpenAI argues the cross-exam mischaracterized a House letter as a Senate inquiry. We have no public record that the court accepted OpenAI's version. We have no public record that the jury was instructed. The public record does not show that yet.

STAN: The court record did give us another piece: Doc. 538, the May eleventh minute entry and trial sheet. It confirms that Monday's trial day ran five hours and forty-seven minutes, with Satya Nadella and Ilya Sutskever called by the plaintiff, Michael Wetter's video deposition played, and Bret Taylor called by OpenAI.

DAVID: The trial sheet is dry, but useful. It authenticates the sequence. Nadella, then Sutskever, then plaintiff rests, then OpenAI calls Bret Taylor. It also shows Microsoft and OpenAI investment materials coming into the record, including a June tenth, twenty nineteen memo from Amy Hood and Kevin Scott to Microsoft's board about the OpenAI investment.

MARCUS: Which leads to Wednesday's live testimony. The Verge reported that Microsoft CTO Kevin Scott took the stand on the final day of testimony. Scott reportedly said Microsoft liked partnering with OpenAI partly because it would show how to build Azure for frontier AI research.

STAN: On cross-examination, The Verge reported, Scott was asked about revenue generated and noted that he was not Microsoft's chief revenue officer. That's not a blockbuster moment. But it fits Microsoft's trial posture: keep the company framed as a commercial partner and infrastructure provider, not the mastermind of OpenAI's mission drift.

MARCUS: Then Mike Wetter appeared for Microsoft. The Verge described him as a corporate-development executive who worked on the twenty twenty-one and twenty twenty-three OpenAI deals. Wetter reportedly testified that Microsoft's due diligence found no conditions related to Elon Musk, and that Microsoft did not talk to Musk during that due diligence process.

DAVID: If credited, that undercuts a certain version of Musk's theory: not the whole case, but any implication that Microsoft's dealmaking was built around hidden obligations to Musk. Again, this is live courtroom reporting, not a transcript.

STAN: The Verge also reported that Musk's lawyers pressed on Microsoft's decision rights. Wetter testified, according to that report, that Microsoft had approval rights on some transactions, but not a majority of the board, and that all limited partners had major decision rights. That's a subtle point, but it matters because the trial keeps circling the same word: control.

MARCUS: Control over OpenAI. Control over AGI. Control through board seats. Control through investment terms. Control through a nonprofit parent. Control through a public-benefit corporation. Everyone in this case says they are worried about control. They just disagree about whose control is dangerous.

DAVID: The final witness thread was OpenAI's Josh Achiam, described by The Verge as OpenAI's chief futurist. He reportedly began at OpenAI as a safety intern in twenty seventeen and told the court the early lab felt like an extension of a graduate student lab: collegiate, academic, highly intellectual, without a publish-or-perish culture.

MARCUS: That testimony is part of OpenAI's softer defense. It tries to remind the jury that the organization was not born as a Microsoft pipeline or a pure commercial machine. It was a small research culture trying to figure out what safe AGI even meant. Whether that origin story survives the for-profit turn is the fight.

STAN: And it connects with Zico Kolter's safety testimony from Tuesday. The Verge reported that the chair of OpenAI's safety and security committee said the committee had formally delayed model releases. That is useful for OpenAI because it says: we still have governance mechanisms that can slow the business down.

DAVID: But it is also useful for Musk's side in another way. If release delays are necessary now, Musk can argue that the stakes of governance were real all along. Safety is not a side issue. The judge has kept the trial from becoming a referendum on AI doom, but governance and safety remain embedded in the facts.

MARCUS: The most interesting part of today's update is that the courtroom drama has become procedural. A cross-examination question turns into a curative-instruction request. A congressional letter becomes an exhibit. A Microsoft due-diligence witness becomes a control witness. This is how high-concept AI governance gets translated into trial mechanics.

STAN: And for jurors, the cleanup motion may cut two ways. OpenAI can argue Musk's lawyers overreached and tried to smear Altman with a congressional implication the document did not support. Musk's side can argue the broader investment and conflict questions are legitimate, even if OpenAI disputes the phrasing.

DAVID: That's why the House letter matters but should not be overstated. It is not a verdict on Altman. It is not a finding that OpenAI misused nonprofit capital. It is a congressional information request, attached to a court filing because OpenAI says Musk's lawyer described it inaccurately in front of the jury.

MARCUS: The exhibit itself asks for a briefing and documents by May twenty-second. That means it may outlive this trial week. Even after closing arguments, OpenAI's governance and conflict-of-interest issues will still be moving in Congress, in reporting, and in the court of public opinion.

STAN: Meanwhile, the trial is nearing the turn from evidence to argument. The Verge described Wednesday as the final day of testimony. If that holds, closing arguments are next. The jury will have to convert a messy origin story into legal questions: promise, reliance, control, remedy, and whether Musk is protecting a mission or relitigating a power struggle.

DAVID: The public docket still lags the courtroom. As of our one p.m. Pacific check, we do not have the May twelfth transcript, the May thirteenth transcript, a ruling on the curative instruction, or a newer minute sheet. We do have enough primary material to say this: Altman's cross-examination did not end when he left the stand. It spilled into filings.

MARCUS: And that filing turns a theme we've followed all trial into a documentable dispute. In the courtroom, Musk's side is asking whether Altman can be trusted. In the filing, OpenAI is asking whether Musk's side can be trusted to describe the evidence fairly.

STAN: That's the shape of the case now. Two trust narratives, both aimed at the jury. Musk says OpenAI broke faith with the mission. OpenAI says Musk broke faith with the facts, with the company, and now, in this filing, with the record.

DAVID: We'll keep watching for the ruling, the transcript, and the first public signs of closing arguments. For now, the careful headline is this: OpenAI has asked the judge to clean up part of Altman's cross-examination, while Microsoft's witnesses and OpenAI's safety witness close out the evidence phase. I'm David Carver. Stay critical.