ELON MUSK RETURNED TO THE STAND Thursday and completed his testimony in the federal lawsuit he brought in 2024 against Sam Altman and OpenAI.
The lawsuit alleges that under Altman’s leadership, OpenAI, the nonprofit San Francisco artificial intelligence developer, has abandoned its charitable mission and has become a primarily profit-focused commercial enterprise in violation of the law that governs nonprofit companies.
The case has attracted widespread attention because OpenAI — founded in 2015 by Musk, Altman, Greg Brockman and Ilya Sutskever — has become one of largest and most valuable AI developers.
William Savitt, OpenAI’s lead lawyer, picked up his cross examination by suggesting that Musk’s trial testimony was different than his sworn testimony with respect to what he read in a 2018 term sheet.
He quickly moved into questions that tested Musk’s tacit acceptance of investments in OpenAI’s for-profit subsidiary in 2019 and 2021. Musk said that those investments were not problematic, in his view, because returns were capped, meaning there was a limit on how much investors could earn. He said capped investments in a nonprofit’s subsidiary were consistent with a charitable mission — unless the limits were set so high they allowed unlimited profits.
Savitt then pressed Musk on his own businesses. He elicited testimony that Tesla, SpaceX, Neuralink and xAI are for-profit businesses and that xAI has not open-sourced its most current models.
Savitt brought out that Musk, together with several for-profit investors, proposed to buy OpenAI’s assets on Feb. 10, 2025. The implication was that, if the bid were successful, Musk intended to operate those assets as a for-profit, not a charity.
A number of questions focused on how Musk was unaware of nonprofit OpenAI’s charitable efforts and therefore his view that defendants had “stolen the charity” — a Musk refrain throughout his testimony — was uninformed.
Musk’s best moment on cross was when Savitt asked him why — if having a nonprofit owner of technology was really so important — had Musk never formed a nonprofit to pursue artificial general intelligence. Musk said forcefully, “I did. I created OpenAI.” He added, “Why would I create another one?”
Microsoft’s cross
Alongside OpenAI, Altman and Brockman, the suit also names Microsoft as a defendant, as Musk claims the company played a role in moving OpenAI to a for-profit structure. (Co-founder Sutskever, however, is not a named defendant.)
Russell Cohen, Microsoft lead lawyer, conducted a short, clear and targeted cross examination.
Microsoft was sued in November 2024 and contends that there is a three-year statute of limitations on Musk’s claim against it for aiding and abetting breach of a charitable trust. Microsoft asserts that if Musk knew that he had a claim against Microsoft before November 2021 and failed to sue within three years, his suit against Microsoft is too late.
Cohen’s questioning led Musk to acknowledge he had previously testified that he was “literal” and “said what he meant.”
Cohen then pulled up a thread of posts on Twitter, including one by Musk on Sept. 24, 2020, a date well before the limitations period allegedly applicable to Microsoft.
The post followed a Sept. 22, 2020, public announcement that OpenAI had given Microsoft an exclusive license to its GPT-3 language model, a predecessor to the model behind ChatGPT.
A thread started by a user identified as @WholeMarsCatalog referenced the announcement, tagged Musk and wrote: “I thought OpenAI was supposed to democratize this tech…not give Microsoft an exclusive license.”
A second poster added, “Exclusive License. Might as well rename it ClosedAI.” Musk then weighed in, posting: “This does seem like the opposite of open. OpenAI is essentially captured by Microsoft.” Cohen seized on the second sentence and pressed Musk to agree that he fully understood — as of Sept. 24, 2020 — that OpenAI had been “captured” by Microsoft.
Musk explained that he meant what he said at the time but was subsequently reassured by Altman that OpenAI remained on mission and was free to release its technology to the public.
After getting Musk to confirm he’d made the statement about Microsoft outside the legal time limit for the claim, Cohen asked nothing further.
AI expert testimony delayed
Musk’s lawyers were expected to call Dr. Stuart Russell, professor of computer science at UC Berkeley, to appear as an expert witness on Thursday. As a prominent figure in the world of AI and an author of a widely adopted textbook on the topic, Russell is known for his assessments of the risks of uncontrolled AI development.
Russell did not begin his testimony Thursday because U.S. District Court Judge Yvonne Gonzalez Rodgers excused the jury early. However, earlier in the day, before the jury entered the court room, the judge and the lawyers had a discussion on the scope of Russell’s testimony.
“This is not a trial on the safety of AI or if AI has damaged society. We are not going to get sidetracked into those issues.”
Judge Yvonne Gonzalez Rodgers
Russell came to court with a pretrial expert report addressing several specific questions about AI safety. OpenAI moved to exclude the testimony as unreliable. The judge denied much of the motion, but she ruled that Russell could not present his opinion on the likelihood of a catastrophic or extinction event.
Steven Molo, Musk’s counsel, had prepared a slide to use during Russell’s presentation that referred to the risk of a catastrophic event. OpenAI objected that it was inappropriate under the judge’s prior ruling.
Molo argued that the judge had only barred testimony that quantified the likelihood of an extinction event in numerical terms — not references to the risk.
OpenAI renewed its objection.
The judge said Russell could not discuss anything about a catastrophic or extinction event. She said the trial was about breach of charitable trust, not about AI safety.
“This is not a trial on the safety of AI or if AI has damaged society,” she said. “We are not going to get sidetracked into those issues.”
His voice rising, Molo said it was ironic that OpenAI “says they are about AI safety” but are trying to block the jury from hearing about the risk of an extinction event.
“It is a real risk,” Molo added.
The judge responded, “I believe that you believe that.” She added, “It is also ironic that your client is developing the same technology.”