Elon Musk spent the better part of three days on the witness stand, accusing OpenAI and its executives of having tricked him into donating money to help found what is now one of the largest AI companies in the world.
The lawsuit pits Musk against his former employees-turned-competitors, OpenAI CEO Sam Altman and President Greg Brockman, who Musk says unjustly enriched themselves by straying from OpenAI’s founding mission as a nonprofit to become a for-profit company. Musk also named Microsoft as a co-defendant in the case, accusing the company of aiding and abetting OpenAI’s breach of charitable trust.
The trial’s big personalities and high stakes were on display in court, with Musk regularly clashing with OpenAI’s lawyer, accusing him of trying to “mislead” me. The judge occasionally reprimanded the parties involved, even telling Musk at one point to actually answer the questions being asked of him and warning them to stop discussing whether AI will cause the extinction of humanity.
OpenAI and Microsoft argued that Musk supported the creation of a for-profit division of the company. They claim that he is only filing the lawsuit because he was unable to take full control of OpenAI and now wants to take down a competitor.
Musk’s AI plans under scrutiny
William Savitt, a lawyer for OpenAI, suggested that Musk left the OpenAI board in February 2018 because he was prevented from taking unilateral control of the company. Musk, however, said he left the board to focus on his other companies, including SpaceX and Tesla.
Savitt suggested that in the years after Musk left the board, he took steps to harm OpenAI, especially after forming a competing company, xAI.
During questioning, Savitt asked whether Musk had revealed that he had founded his own AI company when he signed an open letter in 2023 advocating halting the development of AI systems more powerful than OpenAI’s GPT-4. Savitt also mentioned the Musk-led attempt last year to buy OpenAI with a group of for-profit investors, to which Musk responded: “There’s nothing wrong with having a for-profit organization, you just can’t steal from a charity.”
Savitt also pressed Musk on why he hasn’t created an AI nonprofit since leaving the OpenAI board. Musk said he didn’t create a new one because he had already founded OpenAI.
“Why would I create another nonprofit when I already founded one? That doesn’t make any sense,” Musk said.
Debate on AI security risks
The courtroom debate extended beyond OpenAI’s founding, addressing the security risks posed by AI, just before the interrogation began on Thursday.
“We could all die” because of AI, Steven Molo, Musk’s lawyer, told OpenAI’s lawyer and Judge Yvonne Gonzales Rogers before the jury and Musk took their seats on Thursday. But Judge Gonzales Rogers said such alarmist statements would not be allowed in front of the jury, especially considering that Musk had founded xAI, his own for-profit AI company.
“I suspect there are a lot of people who don’t want to put the future of humanity in Mr. Musk’s hands, but that doesn’t matter, we won’t get into those issues,” Rogers said, noting that the trial isn’t about whether or not AI has harmed humanity.
Musk registered a for-profit company
The central question of the case is whether OpenAI and its executives unfairly turned the company into a for-profit entity, violating its original mission and misleading Musk.
Musk was one of the company’s co-founders and contributed $38 million to OpenAI. However, he stepped aside in 2018 and suspended all payments until 2020.
“I gave them free funding to create a startup,” Musk testified, stating that he thought he was making a donation to a non-profit organization that aimed to develop AI ‘for the good of humanity’.”
But as early as 2015, before OpenAI’s official announcement, Musk had proposed that OpenAI include a for-profit entity, according to emails presented to the jury. In 2017, he instructed his senior advisers to register a for-profit company in OpenAI’s name, OpenAI’s lawyer said, pointing to meeting notes and the registration documents.
Musk testified this week that he didn’t mind OpenAI having a for-profit subsidiary as long as it didn’t “outgrow” the nonprofit, which he said is what ended up happening.
Musk “didn’t read the fine print”
On Wednesday, Savitt showed Musk emails and text messages from 2018 in which Altman tried to inform Musk about OpenAI’s plans to secure additional funding from Microsoft. (Musk did not respond to all messages.)
An email from the time included a draft proposal for a corporate structure that explicitly stated that OpenAI intended to raise $10 billion in the future — but Musk testified that he “didn’t read the fine print.”
“It’s a four-page document,” Savitt responded.
But Musk testified that his trust in OpenAI’s leaders began to shake. Musk told Altman in 2022 that OpenAI’s $20 billion valuation following Microsoft’s $10 billion investment seemed like a “bait and switch.”
“I agree it looks bad,” Altman responded, before noting that Musk turned down the equity stake OpenAI offered him.
At the root of all this is Google DeepMind
Musk’s race to build better AI than Google’s was a motivating factor in his funding of OpenAI, he testified. Google’s DeepMind lab, for example, has been producing significant research for years.
“DeepMind is moving too quickly. I worry that OpenAI is not moving fast enough to keep up. Creating it as a nonprofit may have been, in retrospect, a poor decision,” Musk said in a 2016 email sent to one of his colleagues at Neuralink, another of Musk’s companies.
Musk testified on Tuesday that he was concerned that Google’s approach to AI was not secure enough. There needed to be “some kind of counterpoint” to Google, “an open-source nonprofit as opposed to a closed-source for-profit one,” Musk said.
Musk’s heated exchanges with OpenAI lawyer
OpenAI lawyer Savitt questioned Musk for two days, on Wednesday and Thursday. At times, exchanges between them became tense.
Savitt asked Musk to stick to “yes” or “no” answers, and at one point Musk asked if Savitt could stop interrupting him.
“Your questions are not simple. They were designed to mislead me,” Musk told Savitt early Wednesday, before comparing the question to the classic fallacy of “have you stopped beating your wife?” The judge interrupted Musk, saying they would not “go into that.”
After the jury and Musk left the courtroom on Wednesday, Judge Gonzales Rogers admitted to OpenAI’s lawyers that Musk “was, at times, difficult.”
