Shivon Zilis, the longtime Musk advisor and mother of four of his children, took the stand in the Musk v. Altman trial this week and may have done more damage to Elon Musk’s case than any witness so far. According to The Verge AI, which had a reporter inside the courtroom, Zilis spent hours walking the court through emails, text messages, and contemporaneous notes from 2017 and 2018 that gut Musk’s claim that OpenAI betrayed its founding mission.
This matters because Zilis was the only person taking notes when Musk, Sam Altman, Greg Brockman, and Ilya Sutskever debated turning OpenAI into a for-profit. The Verge AI reports those notes are now the trial’s most important evidence, more important even than Brockman’s diary. What Musk’s team likely hoped would be friendly testimony from a loyalist instead handed OpenAI’s lawyers a roadmap to his own intentions.
What the notes actually show
Zilis testified she worked across Musk’s “entire AI portfolio: Tesla, Neuralink, and OpenAI” starting in 2017, putting in 80 to 100 hour weeks finding bottlenecks. The paper trail she left behind tells a specific story:
- One email floated the option to “switch to for profit in next couple of weeks (woah fast!).”
- Another logged that Altman, Brockman, and Sutskever had a “complete non-negotiable” demand: an ironclad agreement that Musk (or anyone) would not have absolute control of any AGI they built.
- Zilis wrote to Musk’s money manager Jared Birchall: “They say they will not move forward without a guarantee to switch away from him having control.”
- She knew about Musk freezing OpenAI’s funding on August 20, 2017. Musk didn’t tell Brockman and Sutskever until September 1.
- Musk floated putting Zilis, Sam Teller, and Birchall on OpenAI’s board so he’d control the nonprofit.
- After Musk hired Andrej Karpathy, he asked Zilis for a list of top OpenAI people to poach.
One of her brainstormed solutions for Musk: recruit DeepMind’s Demis Hassabis, because “if he hung around E perhaps it would force him to think about humanity more.”
Why this reframes the lawsuit
Musk’s core argument is that OpenAI’s pivot to a for-profit structure violated its mission. Zilis’s own emails show the for-profit conversion was being actively discussed in 2017, while Musk was still in the picture, and that the cofounders’ red line was specifically blocking him from controlling the resulting AGI. That’s a hard story to square with betrayal.
The credibility problem is just as steep. Zilis served on OpenAI’s board while secretly having twins with Musk in 2021, only disclosing the paternity after Business Insider reported it. She told Brockman the relationship was “platonic” and the kids came via IVF. She was originally a plaintiff in this very suit. On cross-examination, OpenAI attorney Sarah Eddy noted that context Zilis offered on the stand didn’t appear in her deposition, prompting the line: “Your long-lost memories have been recovered.”
What it means for the AI industry
The trial is shaping up as the most detailed public look yet at how OpenAI’s structure was negotiated, and it’s landing as every major lab debates governance, control, and conversion to for-profit status. A few takeaways for practitioners and operators watching:
- The “who controls AGI” question was a founding-document-level fight at OpenAI, not a recent invention. Expect that framing to surface in future governance disputes at Anthropic, xAI, and others.
- Internal notes and Slack-equivalent paper trails are now courtroom artifacts. AI leaders should assume every email about structure, funding, or board seats can end up on a screen for a jury.
- Musk’s xAI competes directly with OpenAI today. A loss here weakens his moral-high-ground positioning against Altman, even if it doesn’t change xAI’s product roadmap.
The trial continues, and The Verge AI has more from inside the courtroom at the original report.