Shivon Zilis Testifies in Musk v. Altman Trial
- From AI wunderkind to Musk’s quiet operator
- Inside the OpenAI board — and the ChatGPT blindside
- The 2023 coup and a Microsoft “terrifying” embrace
- Musk’s secret merger fantasy, in her own emails
- Memory holes and cross-examination
- Whose loyalist, whose story?
Shivon Zilis Testifies in Musk v. Altman Trial Human Human coverage presents Shivon Zilis as a pivotal but conflicted witness whose emails, notes, and personal ties expose both governance lapses at OpenAI and Musk’s push to pull its AI work into Tesla. These reports emphasize her mixed loyalties, memory gaps under questioning, and evolving views on Microsoft’s influence, framing the trial as a messy struggle over power, control, and credibility rather than a straightforward case of one side betraying AI safety ideals. @Verge Shivon Zilis walked into the Musk v. Altman courtroom as the ultimate insider: AI investor, former OpenAI board member, Tesla and Neuralink executive, and mother to four of Elon Musk’s children. By the end of the day, her attempt to play neutral referee instead of partisan enforcer had turned into the trial’s sharpest irony: the person Musk may trust most has supplied some of the toughest evidence against him.1
From AI wunderkind to Musk’s quiet operator
In testimony, Zilis sketched a classic Silicon Valley origin story. After Yale, she went to IBM, then Bloomberg Ventures, eventually helping launch Bloomberg Beta, an early AI-focused fund.8 Her obsession with AI, she said, began at 13, when she read Ray Kurzweil’s The Age of Spiritual Machines “10–15 times,” a formative text for a generation of techno-optimists.8
That obsession eventually brought her into Musk’s orbit. She went on to run what she described as his “entire AI portfolio” across Tesla, Neuralink, and OpenAI, working 80–100 hour weeks in what she called a “bananas” job, according to reporting on her testimony.1 When asked in court about Musk’s own workload, she laughed and described him as going into “maniac mode,” with their time together serving as a brief break from “the insanity … that is his entire work life.”9
Privately, their relationship was more than professional. Zilis confirmed under oath that she is the mother of four of Musk’s children, including twins born in 2021 while she was sitting on the OpenAI board — a fact she kept from fellow directors and even her own family until court documents exposed it years later.1 When OpenAI president Greg Brockman confronted her after the news broke, she told him their relationship was “entirely platonic” and that the twins were via IVF, a reassurance that helped her keep her board seat despite other directors wanting her gone.2
Inside the OpenAI board — and the ChatGPT blindside
On the stand, Zilis painted herself as a conscientious board member increasingly uneasy with how Sam Altman ran OpenAI.
Her first major rupture, she testified, came when ChatGPT was unleashed on the world in late 2022 without advance notice to the nonprofit board. Zilis said she had “major concerns” that the board wasn’t told in advance, and that she and the “entire board had voiced extreme concern about that whole massive thing happening without any semblance of board communication.”13 She described this as the first serious concern she raised internally about Altman.13
Her worries didn’t stop there. She told the court she also flagged a proposed deal with Helion, a speculative nuclear energy company in which both Altman and Brockman were investors.11 To Zilis, OpenAI contemplating a major commitment to a company that “didn’t have an official product yet” “felt super out of left field,” and she pressed: “How is it the case that we want to place [a] major bet on a speculative technology?”11 She recalled the Helion situation as one of the rare moments she felt a pit-of-the-stomach alarm strong enough that she explicitly told fellow directors: I’m not okay with this.10
These episodes set up Musk’s narrative: that Altman steered OpenAI away from its altruistic mission and into opaque, conflicted deals behind the backs of the nonprofit board. But as the timeline moved forward, Zilis’ story stopped cleanly serving Musk’s case.
The 2023 coup and a Microsoft “terrifying” embrace
When OpenAI’s board abruptly fired Altman in November 2023, triggering one of tech’s most chaotic weekends, Zilis was no longer on the board but still close to the players. In the moment, she reached out to Altman personally, texting him: “I just wanted to say I hope you are [OK]. I have no idea what’s going on but … I care about you as a person first and foremost. Sending all of my positive vibes your way.”6
Later, as the dust settled and Microsoft swooped in to back a reinstated Altman, Zilis testified that the fallout “changed her view” of the OpenAI–Microsoft alliance.7 She recalled Microsoft CEO Satya Nadella boasting that the company was “below them, above them, around them” in relation to OpenAI — a phrase she interpreted as “complete control” and “terrifying because [it] was just not the thing that we had been fighting so hard for.”7
Zilis said she was disturbed that board members who had voted to oust Altman were effectively “expelled,” and even more alarmed that the law firm hired to investigate the fiasco never publicly explained what really happened.7 Her account fits Musk’s broader claim that OpenAI drifted from its non-profit, safety-first roots into a for-profit machine tightly bound to Big Tech.
Altman, for his part, has pressed on in public as the face of frontier AI safety work, amplifying OpenAI’s technical defenses like “chain of thought monitors” to detect misaligned AI reasoning.15 But Zilis’ courtroom recollections pushed the focus away from model alignment and back toward basic corporate governance: who was really in charge, and who got to know what, when.
Musk’s secret merger fantasy, in her own emails
If that were all Zilis had to offer, her appearance would have been a clean win for Musk. Instead, OpenAI’s lawyers rolled out her own emails and brainstorm docs — and that’s where things turned brutal.
In internal messages, Zilis emerges as a strategist for Musk’s long-standing desire to bring OpenAI under his direct control. “We are seeing more details about Zilis advocating for Musk’s plan to wrap OpenAI into Tesla,” one courtroom liveblog noted, quoting an email where she wrote, “Tesla solves the funding issue immediately… Tesla at least has option to bury.”4 In another, she enthused: “They haven’t internalized the advantages to burying this in Tesla for stealth advantage.”4
A separate document showed her “brainstorming possible scenarios for AI,” most of which put Tesla at the center. One option: “OpenAI as a b-corp subsidiary of Tesla.” Another: “Altman as anchor for TeslaAI.”12 In perhaps the most revealing line, she floated a recruitment move against rival DeepMind: “Find a way to get Demis. Seriously…. Demis really does fanboy hard and I don’t think he’s immoral… just amoral. If he hung around E perhaps it would force him to think about humanity more.”12
For Musk, who presents himself online as the uniquely incorruptible steward of superintelligence — a framing he implicitly endorsed by retweeting Chamath Palihapitiya’s line that “the only person that we can trust is Elon … I feel like he’s the least corruptible”14 — those emails land like a boomerang. They suggest that while Musk now lambasts OpenAI’s for-profit pivot, he once sought to tuck the lab inside his own car company, with the explicit “stealth advantage” of being able to “bury” it on Tesla’s balance sheet.4
That contradiction underpins one Verge analysis that bluntly summarized the day’s events: “Musk’s Biggest Loyalist Became His Biggest Liability.”1
Memory holes and cross-examination
As the documents piled up, the credibility of Zilis’ recollections became a battleground.
OpenAI’s lawyer Sarah Eddy pressed Zilis on why some messages she claimed not to remember in earlier depositions were now suddenly familiar after repeated document reviews. When Zilis said she now recalled certain messages because she had gone over the evidence “numerous times,” Eddy responded with a needle-threading sarcasm: “Your long-lost memories have since been recovered.”5
On another thread, Zilis tried to put distance between herself and a text exchange with Musk about OpenAI’s blockbuster deal with Microsoft, in which she had apparently highlighted that the structure was “not maximum profit” and that Microsoft “was not in control.” Asked to explain, she instead leaned into a quirky formulation: “It’s not in my neurons,” she said, standing in for “I don’t remember.” She acknowledged that she could see the messages on the screen but insisted, “it’s not in my neurons, it’s not in my brain, but I see it.”3
The line, repeated in live coverage, became shorthand for the day’s uneasy dance: a witness trying to appear precise and technical about her own fallible memory, while lawyers on both sides tried to pin her to sharply conflicting narratives.
Whose loyalist, whose story?
Chronologically, Zilis’ testimony traces the larger arc of the Musk–Altman civil war:
- Pre-2020: She’s an AI true believer turned Musk lieutenant, helping run his AI empire after an early OpenAI connection.1
- 2021: She quietly has twins with Musk while on the OpenAI board, hiding the relationship from colleagues and family.1
- 2022–2023: As a director, she grows alarmed by ChatGPT’s surprise launch and by Altman-linked deals like Helion, voicing those concerns internally.1013
- Late 2023: She reacts with personal sympathy to Altman’s ouster, then recoils at Microsoft’s muscular rescue and the effective purge of dissenting board members.67
- Behind the scenes across these years: she helps Musk sketch scenarios to fold OpenAI or its talent into Tesla, praising the ability to “bury” the project on Tesla’s books for “stealth advantage.”412
Both sides now claim her as corroboration.
Musk’s camp leans on her boardroom complaints — about being blindsided by ChatGPT, about speculative conflicts like Helion, about Microsoft’s “complete control” posture — to argue that OpenAI abandoned its founding principles and the nonprofit board that was supposed to enforce them.71013
OpenAI’s lawyers, in turn, brandish her emails and strategy notes to argue that Musk is hardly a disinterested idealist. They depict a would-be acquirer eager to subsume OpenAI into Tesla, where it could be funded, “buried,” and leveraged for competitive advantage — precisely the kind of concentrated power Musk now rails against.14
In the end, Zilis didn’t cleanly resolve who betrayed whom in the breakup of Silicon Valley’s most consequential AI partnership. Instead, she gave the court something messier and more revealing: a portrait of a tight, elite circle where personal relationships, secret children, boardroom governance, billion-dollar deals, and apocalyptic concerns about superintelligence are all hopelessly intertwined.
In that world, “Who can you trust?” isn’t just a podcast line Musk retweets.14 It’s the central question of the trial — and on this record, the answer is that everyone’s trust looks at least a little bit “buried.”
Story coverage
Write a comment