On May 6, local time, the case of Musk v. OpenAI entered a more exciting day. On this day, two very key female witnesses appeared in court. One is Shivon Zilis. She worked at OpenAI and served as its director from 2020 to 2023; now she is a Neuralink executive and the mother of Musk’s four children. Her testimony directly brought to the table the looming hidden line of information between Musk and OpenAI. The other one is Mira Murati. She is the former CTO of OpenAI and also served as interim CEO after Altman was briefly dismissed by the board of directors in 2023. Her video testimony was even harsher, pointing the finger directly at Altman himself: confusion, distrust, inconsistent statements, and even deception.
The two testimonies happened to go in two directions, pointing directly at the credit issues of both parties. Was Musk kept in the dark, and is OpenAI trustworthy?
This drama has reached the most exciting part.
“Keep staying close and let the information flow”
Zellis is no ordinary witness in this case In the lawsuit, her identity is almost inherently explosive: a former OpenAI employee, a former OpenAI director, a Neuralink executive, an important person in Musk's close relationship, and the mother of Musk's four children.
So when she appeared on the stage, everyone paid attention not only to "what does she know", but more importantly: whose side is she on?
OpenAI’s side is trying to prove that Musk is not a former founder who was completely excluded from OpenAI - he did not suddenly discover that OpenAI has deteriorated many years later and then angrily sue. He may have always known what was going on inside the company and even had access to information.
The most dazzling piece of evidence is the text message. Information that emerged during the trial showed that in February 2018 (when Musk was about to leave the OpenAI board), Zelis asked Musk whether he should continue to "stay close" to OpenAI in order to "keep the flow of information."
Her exact words: “Do you want me to stay close to OpenAI so that the information continues to flow, or do you want me to start to drift apart? The trust game is going to get tricky right now, so I would be very grateful if you could guide me on how to be worthy of you.”
Musk’s answer was straightforward: Keep getting close to . But he added: Tesla will actively try to poach three or four people from OpenAI, and more people will come later, but it will not actively recruit.
What’s even more intriguing is that Musk himself admitted during cross-examination that he really wanted to know what happened to OpenAI. Business Insider reported that when OpenAI lawyers asked about the text message that kept Zelis close to OpenAI, Musk replied: "Well, I want to know what happened. "
As soon as this sentence came out, Musk's "victim narrative" was not so clean.
He can certainly say that he is the founder who was betrayed. But OpenAI can also ask: If you have been understanding the internal dynamics of OpenAI through Zelis, can you still say that you are completely kept in the dark?
However, the most interesting thing about Zellis’s line is that it is not a one-way attack on Musk.
Zellis himself does not accept the role of "Musk insider". Public reports show that she denied in court that she delivered OpenAI internal information to Musk, and emphasized that her loyalty is not to serve a certain person, but to make AI benefit mankind.
OpenAI wants to shape her into Musk’s eyes and ears, but she herself refuses to be defined in this way.
Even more subtly, her testimony also stoked the fire back within OpenAI.
Zellis mentioned that he had been worried about OpenAI’s potential deal with nuclear fusion company Helion. The reason is very sensitive: Ultraman and Brockman are both investors in Helion, and the company did not have a formal product at the time.
Zellis said in court that the incident made her feel "very abrupt" and asked: Why would OpenAI make a major bet on a speculative technology?
This detail is very suitable for Musk: If OpenAI wants to say that it has to be commercialized just for the sake of its mission, it must answer another question - have the personal investment, corporate cooperation and governance boundaries of the core management been clearly explained?
Zellis’ position in this lawsuit is a lot like a door that’s ajar. On one side of the door is Musk, on the other side is the OpenAI board of directors.
OpenAI wants to prove that Musk is not the person standing outside the door. He can at least always see the light in the crack of the door. On the other hand, Musk can say that what comes out of the crack in the door is precisely the unclear relationships, interests and governance issues within OpenAI.
Zellis is not the heroine of this scene, but she is like the most dangerous middleman in all suspense films: she may not necessarily have the whole truth, but she knows both sides at the same time, making both sides uncomfortable.
"I can't trust Ultraman's words"
If you say Zelis's testimony put the private information line between Musk and OpenAI in court, and Mulati's video testimony directly opened up the trust cracks within OpenAI for us to see.
Meera Murati is not an ordinary employee who has resigned.
She is the former CTO of OpenAI. She has been deeply involved in the development and release of core products such as ChatGPT and DALL·E. She was also one of the most important technical leaders in the years when OpenAI moved from a laboratory to a global AI center.
In November 2023, after Altman was suddenly dismissed by the board of directors, Mulati was also appointed as interim CEO, briefly standing at the forefront of the OpenAI power center. Later, Altman was reinstated and she returned to the CTO position; in September 2024, she announced her departure from OpenAI on the grounds that she wanted to create time and space for her own exploration; after that, she founded a new AI company, Thinking Machines Lab.
On May 6, Mulati’s video testimony was played in court.
She said, Altman created "confusion" and distrust among OpenAI executives, saying different things to different people and sometimes acting deceptively.
OpenAI’s core line of defense in this lawsuit is: commercialization is not a betrayal of the mission, but to keep the mission alive. But Mulati’s testimony directly pointed out: If the highest levels of the company do not trust each other, then how can it convince the outside world that it can still firmly adhere to its mission of “benefiting all mankind”?
The Verge revealed a specific detail. Mulati said that Altman told her that the OpenAI legal team had approved a new model to bypass the safety committee review because the model did not require deployment safety committee approval. But she later asked OpenAI’s general counsel and found out that this statement was not true.
What OpenAI has always said to the outside world is to "ensure that AGI benefits all mankind"; what it needs the outside world to believe most is that it can not only make the strongest models, but also manage these models safely, transparently and responsibly. But now, the former CTO said in court that he had gone to verify the CEO's statement due to security review issues, but found that it was not correct.
This is not just a management style issue, it is a trust issue.
If someone who is completely on Musk’s side criticizes Altman, the outside world can also say that this is a litigation strategy. But not Mulati. She was pushed to the position of interim CEO by OpenAI and continued to serve as CTO after Altman was reinstated. Her relationship with OpenAI is not that of an outsider, but of the former core layer.
This is also the easiest place to use Musk Square. Mulati’s testimony just gave Musk a good fulcrum: OpenAI, a company that claims to protect the future of mankind, has internal opacity, distrust, and governance risks.
In other words, Mulati’s video testimony pushed the question from “Has OpenAI betrayed its original intention” to an even more unpleasant point:
If OpenAI cannot even fully trust its former CTO in the CEO, how can it convince the public that it can manage AGI on behalf of all mankind?
By the way, Mulati is not the only OpenAI executive who has been exposed to be at odds with Ultraman. Previously, OpenAI CFO Sarah Friar also expressed concerns about Altman’s plan to advance the IPO as soon as the end of 2026.
Altman may be one of the best storytellers in Silicon Valley, but no matter how beautiful the story is, it cannot replace the facts. His surroundings are not monolithic either.
It’s getting uglier and more exciting
If we talk about the opening statement stage, this lawsuit is like a Silicon Valley version of a commercial blockbuster: Musk accused OpenAI of betraying its original intention, and OpenAI counterattacked and Musk failed to gain control.
Then on May 6, it suddenly turned into another kind of drama: "Who is in contact with whom in private", "Did you know what happened inside through her?", "Why can't you trust the words of the CEO"...
This is very much like a palace fight drama. The most tense thing in the play is often not who the emperor favors today and who he deposes tomorrow, but that everyone in the palace knows a little secret, everyone holds a thread, and everyone is both a chess piece and may be evidence for others.
Zelis was like a door that was not closed tightly, and each side could see a little light coming from the other side; Mulati was like a person coming out of the inner hall, revealing that the inside of the palace wall was not stable at all.
The testimonies of the two people are like two knives, one slashing at Musk and the other stabbing OpenAI. Musk wanted to play a founder who was betrayed, but Zelis’s text message made his script muddy; OpenAI wanted to play an idealist who was forced to go commercial, but Mulati’s testimony also made it take off its makeup.
At this point, the lawsuit has turned into something even more embarrassing: everyone is trying to undermine the other party. On the stage, everyone talks about the overall situation, but off the stage it is all about information, relationships, control and boundaries of interests.
Everyone was talking about justice, but what the court finally found out were all old accounts.