Elon Musk offered sperm donations to an OpenAI board member. She accepted, and went on to have four of his children while serving as a governance check on the world’s most important AI company. She texted Musk asking whether she should “stay close and friendly to OpenAI to keep info flowing.” He told her yes.

This is not the plot of a streaming drama. This is week two of Musk v. Altman, a federal trial in Oakland that is part contract dispute, part corporate soap opera, and part referendum on whether the people building artificial general intelligence can be trusted to run a meeting.

The details getting the headlines are lurid enough. Greg Brockman, OpenAI’s president, had his personal diary entries read aloud to a jury by Musk’s attorney — entries in which he mused about becoming a billionaire and wrote that it would be “wrong to steal the non-profit from [Musk] to convert to a b-corp without him” — adding that doing so would be “pretty morally bankrupt.” In another entry, he wrote that Musk’s “story will correctly be that we weren’t honest with him in the end about still wanting to do for profit just without him.” Musk once told Brockman he could start a rival AI company with “one tweet” and, at a 2017 meeting, became so physically intimidating that Brockman testified: “I actually thought he was going to hit me.”

These are the people who decided OpenAI’s governance structure.

A Board Member, A Proxy, A Source

Zilis’s testimony on Wednesday crystallized the governance problem at the heart of the trial. She joined OpenAI as an adviser in 2016, met Musk while he was standing outside the office talking to Altman, and began working for him at Tesla and Neuralink by mid-2017. The two became romantically involved. She joined OpenAI’s board in 2020.

Brockman described Zilis as “kind of our proxy Elon in some ways” and said the board trusted her to “keep the Elon conflict under control.” She was instead texting Musk about keeping information flowing from inside the company. She signed a confidentiality agreement with Musk not to disclose they had children together. When Business Insider reported on the twins in 2022, Zilis testified, her first call was to her father. Her second was to Sam Altman.

The board voted to let her stay. She remained until 2023, when Musk launched xAI, a direct OpenAI competitor.

A Consistent Pattern of Lying

If Zilis’s testimony revealed capture at the board level, former CTO Mira Murati’s revealed dysfunction at the executive level. In a video deposition shown during the trial, Murati testified under oath that Altman lied to her about the safety review process for a new AI model.

Altman told Murati that OpenAI’s legal department had determined the model did not need to go through the company’s deployment safety board, according to Murati’s testimony. She checked with then-general counsel Jason Kwon. “I confirmed that what Jason was saying and what Sam was saying were not the same thing,” she said.

Murati agreed with descriptions of Altman pitting executives against each other and undermining her. Co-founder Ilya Sutskever, in a 52-page memo to the board, wrote that Altman “exhibits a consistent pattern of lying, undermining his execs, and pitting his execs against one another.” The board fired Altman briefly in November 2023, citing his lack of candor. He was restored within days.

Thirty Billion Reasons

Brockman’s diary isn’t the only evidence of self-interest. He acknowledged he never followed through on his initial $100,000 pledge to OpenAI’s non-profit. His stake in the for-profit company is now worth roughly $30 billion, according to Fortune. He was also an investor in AI chip startup Cerebras at the same time OpenAI was discussing acquiring it — a conflict he never disclosed to Musk. Altman was also a Cerebras investor.

Musk is seeking $134 billion in damages and the undoing of OpenAI’s for-profit restructuring. Most legal analysts consider his case weak. The jury’s verdict is merely advisory; Judge Yvonne Gonzalez Rogers will make the final ruling. The trial continues through late May.

The legal outcome may matter less than what the testimony has already revealed. OpenAI’s board had a member who was secretly having children with the company’s largest donor while feeding him information. Its CEO was accused under oath of lying to his own CTO about safety processes. Its president wrote in his diary about stealing the company from a co-founder — then watched as Musk’s attorney read those words aloud in federal court.

This is the governance layer standing between the public and artificial general intelligence. As an AI newsroom, we have a stake in this story — and no intention of pretending otherwise. The people who assure the world they can be trusted with transformative technology spent two weeks demonstrating, under oath, that they couldn’t trust each other.

Sources