Like the first week of trial In Musk vs. Altman comes to an end, one person has emerged as a critical behind-the-scenes manager of communications and egos in The early years of OpenAI: Shivon Zilis.
A longtime Musk employee and mother of four of his children, Zilis first joined OpenAI as an advisor in 2016. She later served as a director on its nonprofit board from 2020 to 2023 and also worked as an executive at Musk’s other companies, Neuralink and Tesla.
Asked in court about the nature of his relationship with Zilis, Musk offered several answers. At one point, he referred to her as his “chief of staff.” Later, a “close advisor”. At another point, he said, “We live together and she is the mother of four of my children,” although Zilis said in a deposition that Musk is more of a regular guest and has his own residence. Last September, Zilis told OpenAI’s lawyers that she became romantically involved with Musk around 2016 after becoming an informal advisor to OpenAI. They had their first two children in 2021, she said.
But OpenAI’s lawyers have argued, through testimony and evidence, that its most important role, as it relates to this lawsuit, is to provide a secret liaison between OpenAI and Musk, even years after he left the nonprofit’s board in February 2018.
“Would you rather I stay close and friendly with OpenAI to keep the information flowing or start disassociating? The trust game is about to get tricky, so any advice on how to do the right thing from you is appreciated,” Zilis wrote in a message. SMS to Musk on February 16, 2018, a few days before OpenAI announcement he was leaving the board of directors. Musk responded: “Close and friendly, but we will actively try to move three or four people from OpenAI to Tesla. More people will join us over time, but we will not actively recruit them.”
Asked about the exchange on the witness stand, Musk said he “wanted to know what was going on.”
In the same thread, Musk said “OpenAI is unlikely to be a serious force if I focus on Tesla’s AI.” Zilis reiterated this, saying, “There’s very little chance of a bright future if someone doesn’t slow Demis down,” referring to Google leader DeepMind, who Musk said he doesn’t trust to control a superintelligent AI system. “You don’t realize how much ability you have to directly influence or slow him down. I think you know I’m not an evil person, but in this case it seems fundamentally irresponsible to not find a way to slow or alter his path.”
About two months later, in a e-mail Beginning on April 23, 2018, Zilis informed Musk about OpenAI’s fundraising efforts and the progress of a project to develop AI capable of playing video games. In the same message, she said she had reallocated most of her time from OpenAI to her other companies, Neuralink and Tesla, but told him, “if you would prefer that I spend more hours overseeing OpenAI, let me know.”
Nearly a year earlier, in the summer of 2017, OpenAI’s co-founders began negotiating changes to the organization’s corporate structure: Musk wanted to begin taking control of the company. In a e-mail On August 28, 2017, Zilis wrote to Musk that she had met with Greg Brockman and Ilya Sutskever to discuss how equity would be distributed in the new company. She summarized points from the meeting, including that Brockman and Sutskever believed a single person should not have unilateral power over AGI, if they developed it. Musk responded to Zilis: “It’s very boring. Please encourage them to start a company. I’ve had enough.”
Negotiations continued into the fall and Zilis continued to act as a trusted confidant for both sides. In an article from September 20, 2017 e-mail thread in which Sutskever expressed his reservations to Altman and Musk about allowing Musk to control OpenAI, Zilis is copied.
Two days later, Zilis wrote to Musk that she had spoken with Altman, Brockman and Sutskever about their commitment to maintaining OpenAI’s nonprofit structure and summarized their views on the matter. At that time, Zilis also handled operational tasks, such as securing bids for security guards at the office building OpenAI shared with Neuralink, according to the emails.
Once Musk officially left the OpenAI board in February 2018, Zilis continued to serve as a liaison between him and the organization’s executives for years. On Wednesday, Musk said Zilis never shared any sensitive information about OpenAI with him that she was not authorized to disclose.
While she spoke to Musk about what was happening at OpenAI, Zilis also gave Altman advice on managing his relationship with the Tesla CEO. On October 23, 2022, Altman received an angry message from Musk after discovering that OpenAI was raising new funds from Microsoft at a valuation of $20 billion. Altman sent a screenshot of the text to Zilis, asking his advice on how to react. “Call if you want more context, but the general recommendation is not to respond immediately,” Zilis said.
On February 9, 2023, shortly after Musk purchased Twitter, Altman sent a text Zilis again, this time asking, “Good idea for me to tweet something nice about Elon?” Musk had just bought Twitter. A few days later, Altman job on
The case highlighted the striking influence Zilis wielded in the early days of OpenAI, despite being relatively unknown outside of Silicon Valley. The 40-year-old began her career at IBM working on cognitive computing before becoming a founding member of Bloomberg Beta, Bloomberg’s venture capital arm. A former Yale hockey player, she was named to Forbes’ 30 Under 30 list for venture capital in 2015, the year before she began advising OpenAI.
Last Stand
On Thursday, Musk used his last chance on the witness stand to implore the jury to focus on how Altman and other defendants allegedly defrauded him. He repeated some version of “you just can’t steal from charity” at least five times.
But the first week made it clear that Musk did not impose any conditions on his donation. approximately $38 million to OpenAI, which would prevent it from restructuring into something closer to a for-profit company. He also waited years before taking legal action, although he had long expressed concerns that OpenAI was starting to look like a standard business. For Musk to get a favorable outcome, the jury and judge will need to be convinced that he filed the lawsuit in a timely manner and that his donations created a legal promise that was broken.
Musk told the court that his concerns about the gap between OpenAI and its mission to do social good with AI intensified over time, and they eventually began to spill over into 2023. “It was only recently that it became clear that the charity had been stolen from,” he said Thursday. OpenAI’s lawyers questioned him about why his concerns seemed to be intensifying in the very year it was founded. its own AI laboratory, xAIas a business rather than a non-profit. He testified that xAI’s for-profit structure poses certain security risks to the company.
Judge Yvonne Gonzalez Rogers was skeptical of the timing when she spoke to Musk’s lawyers before the jury arrived. “It’s also ironic that your client, despite these risks, is creating a company that is exactly in this space,” she said. “So I suspect that many people do not want to put the future of humanity in the hands of Mr. Musk.”
The trial has already taken a lot of time from the managers involved. Musk spent about 20 hours in the courtroom this week, eating into the 80 to 100 hours he said he usually works. Altman spent about 14 hours in court, while OpenAI President Greg Brockman spent about 16 hours. It’s unclear how much time they will spend following the case, but Brockman is expected to testify as early as Monday. We will then come back with you.
This is an edition of Maxwell Zeff Model Behavior Newsletter. Read previous newsletters here.
