Whose Charity Was It, Anyway?

The trial that opened this morning in Oakland is being covered, mostly, as a grudge match. Two billionaires, one of whom used to fund the other, fall out over money and ego, and the falling-out gets dressed up in the language of mission betrayal. Elon Musk wants $150 billion and Sam Altman's job. Sam Altman's lawyers want the jury to see Musk as a jilted investor with a competing AI company and a long memory. The press has its frame ready: Succession with GPUs.

I want to suggest a different frame, because I think the one we're being handed is wrong in a way that matters beyond this courtroom.

Musk v. OpenAI is not really a fight about whether Sam Altman was a good fiduciary of Elon Musk's donations. It is a fight about what happens when a charitable structure, chartered to develop a transformative technology "for the benefit of humanity", is converted, in stages, into a capped-profit subsidiary of a trillion-dollar cloud company, and then, as is happening this month, restructured again into something closer to a conventional for-profit with the nonprofit retained as a minority shareholder and a vestigial conscience.¹ The question on trial is who gets to decide that this conversion was legitimate. The answer the court is likely to give, narrow, contractual, focused on what Musk was promised and whether he got it, will leave the larger question completely untouched.

That larger question is the one I think the public has standing to ask, even if Musk does not.

What was given away

OpenAI was incorporated in Delaware in December 2015 as a 501(c)(3) public charity. Its founding charter committed it to developing artificial general intelligence "in the way that is most likely to benefit humanity as a whole, unconstrained by a need to generate financial return."² Donors gave roughly $130 million in the first three years on those terms. The IRS granted tax-exempt status on those terms. Researchers joined on those terms, several of them have since said, on the record, that the mission language was material to their decision.³

In 2019, OpenAI created a "capped-profit" subsidiary, OpenAI LP, in which investors could earn returns up to 100x their investment before excess profits flowed back to the nonprofit. Microsoft committed $1 billion that year, then $10 billion in 2023, then a further restructuring in 2024 that gave Microsoft expanded rights to OpenAI's models and a substantial equity-equivalent stake in the new for-profit entity.⁴ The cap, OpenAI explained at each stage, was the thing that preserved the mission. Profits above the cap would still go to humanity. The board would still be a nonprofit board. The structure was unusual but the spirit was intact.

The restructure being finalised this month, the timing of which, as the research brief notes, is not coincidental to the trial, removes the cap.⁵ What was a charity that owned a capped-profit subsidiary becomes a public-benefit corporation in which the original nonprofit holds equity. The PBC form is a real legal structure with real fiduciary obligations to a stated mission, and I don't want to be glib about it. But it is a structure designed to let a company pursue profit while considering a public benefit. It is not a structure designed to subordinate profit to a mission. The difference is not rhetorical.

So the sequence, stripped of press releases, is this. Money and labour were gathered under one set of promises. The promises were progressively loosened. Each loosening was justified by the need to raise the next round of capital required to pursue the original mission. And at the end of the sequence, the original mission's institutional vehicle, the charity, is a minority shareholder in a company whose new investors have every reason to want it to behave like a normal company.

I have been trying to think of a clean parallel, and the closest I can get is a hospital that was endowed as a charity, took on private investment to expand, and then was reorganised as a for-profit health system in which the original charitable trust holds a stake. We have rules for that. State attorneys general get involved. There are public hearings. The question of whether the conversion preserves the charitable purpose is litigated, sometimes for years, and the answer is sometimes no.⁶ The conversion of OpenAI is a larger transaction than most hospital conversions and has received substantially less public scrutiny.

Why this isn't really Musk's fight

The reason Musk's case is awkward is that he is not a good plaintiff for the question I think actually matters. He is a donor who left the board in 2018 after, by several accounts, trying and failing to install himself as CEO.⁷ He now runs xAI, a direct competitor. His $150 billion damages claim and his demand for Altman's removal are the moves of someone who wants control, not someone who wants the mission preserved. If he won outright, the most plausible outcome is not that OpenAI returns to its founding charter. It is that Musk gets paid and xAI gets a competitor weakened.

I find it hard to be sentimental about Musk's standing here. But the inadequacy of the plaintiff is not the same as the inadequacy of the question. And the question, was a public-interest commitment converted into private value, and if so, who pays, is one that the trial will only glancingly address, because the trial is bounded by what Musk can claim he was personally promised.

The people who could ask the question with cleaner standing are the state attorneys general of California and Delaware, who do have authority over charitable conversions. Both have been examining the restructure; neither has, as of this filing, moved to block it.⁸ The IRS could ask it, in principle, by re-examining tax-exempt status. The original donors, other than Musk, could ask it, but most have either declined to or are now employed by entities with reasons not to. The researchers who joined on mission grounds could ask it, and a few have, in resignation letters and op-eds, but they are not parties to anything.⁹

This is the structural problem. A charity that holds a transformative technology in trust for the public has, in practice, almost no one with standing to enforce the trust except a litigious ex-donor with his own AI company. That is not a feature of the system. It is a hole in it.

What "for the benefit of humanity" was supposed to mean

I want to be careful here, because there is a version of this argument that collapses into the claim that OpenAI should have stayed a pure research nonprofit and never raised commercial capital. I don't believe that. Training frontier models costs what it costs. The people who founded OpenAI in 2015 did not anticipate that scale, and when they hit it, they had a real choice: raise tens of billions in commercial capital, or fall behind labs that would. The capped-profit structure was a serious attempt to take commercial money without becoming a commercial company. I take that seriously.

But "for the benefit of humanity" was always doing two kinds of work in the OpenAI charter. One was about safety: don't build AGI in a way that gets people killed. The other was about distribution: don't build AGI in a way that concentrates the gains in a few hands. The first commitment is the one that gets discussed. The second is the one that has quietly evaporated.

When OpenAI was a charity, the upside of its work, if the work succeeded, accrued, in principle, to a public-benefit purpose. When it became a capped-profit, the upside accrued to investors up to 100x and then to the public-benefit purpose. When it becomes a public-benefit corporation with the nonprofit as a shareholder, the upside accrues to investors and employees, with a minority claim retained by an entity whose governance the new investors will, reasonably, want to constrain. At each step, the share of the upside reserved for "humanity", which is to say, for purposes other than enriching the people building and funding the company, has shrunk.

This matters because of what the technology is. If OpenAI succeeds at anything close to what it says it is trying to do, the resulting capability will reshape labour markets, displace workers, restructure entire industries. Some of those effects are already visible.¹⁰ The original charter was a promise that the gains from that reshaping would be distributed in some way that took the losses seriously. The current structure is a promise that the gains will be distributed the way gains in tech companies are normally distributed: to founders, employees, and investors, with whatever the public gets coming through normal channels, taxes, products at market prices, philanthropy at the donors' discretion.

That is a much weaker promise. It might still be the right promise, reasonable people can argue that the original one was naive, that capped-profit was unworkable, that the PBC form is the most you can ask of a frontier lab in a competitive market. But the argument has not really been had in public. It has been settled, transaction by transaction, in negotiations to which the public was not a party.

What the trial will and won't do

The Oakland jury will decide a narrow question: did Altman and OpenAI's board defraud Musk specifically, by inducing his donations under false pretences. The advisory verdict expected in May will turn on contracts, emails, and what Musk was told in 2015 and 2016. Whatever the jury finds, the restructure will close. Microsoft will get what it has paid for. The PBC will launch. xAI will continue to compete. Altman will keep his job or, less likely, lose it; either way, the structural conversion is essentially done.

The thing the trial will not do is establish a precedent for how charitable AI labs can convert into for-profit AI companies, what the public is owed when they do, or who can stop them if the conversion is not in the public interest. That precedent is being set, but it is being set by acquiescence, by the absence of regulators willing to intervene, the absence of a clear legal framework for charitable conversions of AI assets, and the absence of any constituency with both standing and motivation to fight.

I keep coming back to the fact that the only person with the resources and motivation to litigate this is the founder of a competing lab. That is not justice catching up with the conversion. That is the conversion being challenged by someone who wants a different cut of the same deal.

The question worth carrying out of this trial, regardless of the verdict, is whether we have any mechanism, short of waiting for a wronged billionaire, for asking what the public was promised by the AI labs that took charitable status, and what, exactly, we are getting in return. Right now, on the evidence, the answer is no. We have a courtroom in Oakland, and after that we have whatever the labs decide to tell us about what they decided among themselves.

That isn't a charitable mission. It's the residue of one.

Footnotes

¶