Trial reaches halfway mark

Week 2 analysis

AS THE LAWSUIT brought by Elon Musk against Sam Altman, Greg Brockman, OpenAI and Microsoft reaches its halfway point, it is worth reviewing how the parties are doing. The suit is based on Musk’s claim that, with help from Microsoft, OpenAI abandoned its charitable mission under Altman and Brockman’s leadership.

OpenAI raises doubts about Musk’s intent

The core strategy for the OpenAI defendants is to keep the focus on Musk, his credibility and his motives — both those announced and unannounced. And if it is true that the best defense is a good offense, OpenAI must be pleased with the state of play. Musk’s broad claim that Altman and Brockman “stole the charity” has gotten bogged down in the granular focus on the wide-ranging and intense discussions among the four founders of OpenAI over a six-week period in August and September of 2017.

At that point the founders — Musk, Altman, Greg Brockman and Ilya Sutskever — were trying to decide what they should do to facilitate raising the hundreds of millions or billions of dollars needed to obtain “compute,” the tech slang for computer processing time and capacity. They needed enormous amounts of computing power to reach the goal of creating artificial general intelligence, or AGI, for the benefit of humanity — the stated nonprofit mission of OpenAI. As a nonprofit, OpenAI had been funded by donations, including $38 million from Musk, but charitable fundraising was proving slow and insufficient for raising the large sums needed.

Through numerous meetings and emails, texts and conversations, the founders debated a variety of different ideas and corporate structures that would facilitate the fundraising. Shivon Zilis and Sam Teller, close advisors to Musk, were involved as well, facilitating communication among the founders.

Among the many ideas discussed was the conversion of OpenAI from a nonprofit to a for-profit corporation, the creation of a for-profit subsidiary of the existing nonprofit, the combination of OpenAI with Musk’s Tesla, and even an initial offering of a new cryptocurrency.

While none of the ideas was adopted at that time, the discussions gave OpenAI’s trial team plenty of opportunity to show or suggest that Musk was no bright shining knight devoted solely to the Camelot of a nonprofit enterprise developing AGI for humanity. (The for-profit subsidiary was formed two years following the discussions.)

Through friendly questioning of Brockman, OpenAI’s lawyer Sarah Eddy was able to make several points. Among them: 1) Musk was not adverse to the creation of a for-profit subsidiary for OpenAI. 2) He was not absolutely against a for-profit subsidiary issuing equity to the other founders as long as his share was larger to reflect his contributions. 3) If there was going to be a for-profit corporation, Musk wanted absolute control of it. And 4) Musk was also serious about solving the fundraising issues by combining OpenAI with Tesla in some way.

Brockman’s testimony also demonstrated that by the end of the discussions, the situation among the four founders was not all love and kisses. Zilis, who shares four children with Musk via in vitro fertilization, later called the founders’ split a “weird half breakup.” Musk had thrown his weight around to kill a proposal, supported by the others, to create a for-profit subsidiary and issue equity for the partners because the proposal did not give him majority control.

In the end, none of the proposals worked. Four people — considered to be among the smartest people on the globe — could not reach agreement with each other. As a result, they defaulted to the structure they started with: OpenAI as a nonprofit still needing enormous amounts of cash.

Musk got the others to commit to retaining the nonprofit structure, saying it was either that or part ways. However, Brockman and Sutskever would not agree to stay with OpenAI for two years and not poach employees after they left.

It was a détente, rather than an agreement.

OpenAI’s lawyers also made hay from emails and texts that suggested Musk was interested in recruiting OpenAI employees to work at Tesla.

Taken altogether, the testimony was effective in raising doubts that Musk was a visionary primarily dedicated to the protection of humanity.

Moreover, OpenAI did a good job of showing that Musk continued receiving information from Zilis about what was going on at the company, at least up until the time she joined OpenAI’s nonprofit board in 2020, and perhaps afterward. Also effective was the evidence that Zilis, Musk’s close advisor, voted in favor of the 2023 transaction where Microsoft invested $10 billion in the for-profit subsidiary and received additional intellectual property rights. That is the very transaction that Musk has characterized as “stealing the charity.”

At the halfway mark, it seems fair to say that OpenAI has done an excellent job of bringing out Musk’s complicated motivations.

Musk focuses on OpenAI’s mission drift

But it was far from a one-way street. If the 2017 discussions were a founders’ fight, the ending was not an agreement, just a standoff.

The close focus on the founder discussions in 2017 also showed that Brockman and Sutskever were not only interested in creating a for-profit subsidiary to raise money for compute and to support recruiting top engineers; they want equity for themselves. Brockman’s personal journal revealed that, as he weighed the various proposals in that period, he was thinking how he could get to $1 billion for himself. He testified that it seemed a fair outcome for his investment of blood and sweat and tears, only to have to explain that what he has received (so far) is a stake now worth close to $30 billion. Steven Molo, Musk’s lead lawyer, battered Brockman with questions about why he had not donated back the excess $29 billion.

Later in the week Musk’s team presented testimony from Helen Toner and Natasha McCauley — two OpenAI board members who voted to fire Altman as CEO in November 2023 only to be forced to reinstate him when almost all of OpenAI’s employees signed a letter urging that Altman be rehired (after Microsoft offered to hire them all).

Toner and McCauley testified that Altman was untrustworthy, withheld information from the board and sometimes lied. As a result, they didn’t feel they could fulfill their function of overseeing the for-profit corporation. Both resigned after he was reinstated.

They discussed a situation that came up earlier in 2023 when Altman falsely told others that the in-house legal team said that ChatGPT-4 Turbo did not require a safety review. While the release of that model did not cause safety issues, they said it demonstrated a subordination of safety to the demands for commercial success. As the technology grew more powerful, they warned the consequences of putting profits over safety could be disastrous.

David Schizer, former dean of Columbia Law School, gave expert opinions on questions that went to the core of whether OpenAI had fulfilled its obligations as a charitable enterprise. Presented with hypothetical questions based on the matters that Musk is seeking to establish through other witnesses, Schizer testified again and again that Altman’s conduct did not fulfill his duties to the board and the charitable enterprise.

Schizer explained that a for-profit subsidiary of a nonprofit corporation is perfectly appropriate if designed to advance the nonprofit’s mission, but not when used as a vehicle to create wealth and profit for insiders.

Schizer was an effective witness for Musk and left the stand relatively untouched by an aggressive cross examination.

Microsoft lies low

Microsoft continued in its strategy of lying low and letting Musk and Altman slug it out. Microsoft is still pressing the story that it came to the dance after the controversial for-profit was in place. Microsoft’s position was that all the company did was provide capital and help OpenAI achieve its dreams.

Schizer’s testimony added another dimension to that story. He used charts to show how Microsoft’s stake in both OpenAI’s value and technology had expanded dramatically in transactions in 2023 and 2025, so that today its stake is worth more than the nonprofit’s.

Microsoft pointed out that Schizer was basing his opinions on facts that still had to be proved. Microsoft’s version of reality will be tested Monday when its CEO Satya Nadella takes the stand.

Can this end well?

When we began our trial coverage, we posed the question: “In a jury trial that might decide the fate of all humanity, should you root for the world’s least sympathetic plaintiff or most devious defendant?”

For many, the question remains unresolved, but that is as it should be.

It was a trick question.

It is hard to root for either side. The only path that gets to a good outcome — and perhaps there is such a path — is to root for the judge.

If the jury comes back with a finding that the defendants (with or without Microsoft) breached the charitable trust, and also that Musk did not wait too long to bring suit, then the “remedies” will be determined, not by the jury, but by U.S. District Court Judge Yvonne Gonzalez Rogers.

Judge Yvonne Gonzalez Rogers is in a position to do something that OpenAI’s founders could not do themselves. Whatever she does, it will probably not make any of the parties happy.

If it gets to that point, Gonzalez Rogers will have a lot of options.

Musk is asking her to impose two different sorts of damages.

First, economic damages, that is the “unlawful gains” that have been received by the defendants. Musk’s expert thinks that up to $109 billion should be “disgorged” — the legal term for when ill-gotten gains have to be coughed up by the wrongdoer. Musk says that he doesn’t want any money for himself; he says that it should go to the nonprofit and be used to pursue the original mission.

Second, Musk seeks a broad suite of “equitable remedies,” including an injunction ordering defendants to operate in accordance with the nonprofit’s mission. He also wants Altman and Brockman out, Microsoft to pay back its gains (part of the disgorgement), and in a real showstopper, he is asking for an unwinding of everything necessary to return the structure to one consistent with the nonprofit’s mission.

Both sorts of remedies fall within the “equitable” jurisdiction of the court. Equity is a body of law that came over to the states from England in times long past and allows a court to issue non-monetary remedies when ordinary monetary damages will not get to a result that is fair and equitable. (Gonzalez Rogers has already decided that disgorgement is an equitable remedy.)

What is fair and equitable is decided by the judge who has very broad discretion. That means that if the jury makes the findings — that the defendants are liable and Musk’s suit was timely — the case will go to Gonzalez Rogers to determine remedies. She will have the power to do what she thinks is best.

Gonzalez Rogers is a hardworking and experienced federal judge, paid by the government and appointed for life. She is in the right place and at the right time to do something that the founders could not do themselves.

She will be able to consider — without the venality and greed that fell upon the four founders — how best to fulfill the mission that is expressed in OpenAI’s nonprofit charter.

Whatever she does, it will probably not make any of the parties happy.

Coming events

Two key witnesses are scheduled early next week: Microsoft’s Satya Nadella and OpenAI co-founder Sutskever, who is not a named defendant in the case. The current schedule involves completing witness testimony on May 13 with closing arguments the following day. Friday is an off day so the case would likely go to the jury Monday, May 18.

The judge has said she wants the parties to begin to present their case on damages on May 18, even as the jury is deliberating. It is unclear what will happen to that process when the jury returns its verdict. The jury functions in an advisory capacity, so technically its findings are not binding on the judge, though she has told the lawyers that she is likely to follow the jury’s lead.