OpenAI’s Founding Myth Is the Real Defendant

Every fast-growing industry wants a creation story. OpenAI’s is now evidence.

As of May 9, 2026, a federal jury trial in Oakland is testing Elon Musk’s claim that Sam Altman, Greg Brockman, Microsoft, and OpenAI-related entities took a nonprofit project built for the public benefit and turned it into a profit-driven machine aligned with one of the world’s most powerful technology companies; the Northern District of California’s public case page summarizes Musk’s complaint as alleging that he was induced to help found and fund OpenAI as a nonprofit devoted to safe AI and broad sharing before the organization shifted toward for-profit structures that allegedly concentrated benefits with Microsoft and affiliated entities (U.S. District Court, Northern District of California¹). OpenAI’s answer, in public and in court, is that Musk is trying to damage a rival after leaving OpenAI’s board in 2018 and building his own AI company, xAI; Reuters reported that OpenAI has portrayed Musk as motivated by a desire for control and by frustration over OpenAI’s success (Reuters via Investing.com¹¹).

I do not think the cleanest version of Musk’s case is that commercialization itself was forbidden. That is too simple. The better claim is sharper: OpenAI gained trust, talent, donations, and moral authority by presenting itself as a nonprofit answer to private capture, then built a structure in which that mission became dependent on the commercial engine it was supposed to discipline. On that point, I think the betrayal story is stronger than the adaptation story.

Start with the promise. OpenAI’s 2015 launch announcement did not sound like a normal venture-backed startup with a noble mission statement taped to the lobby wall. It described OpenAI as a nonprofit artificial intelligence research company whose goal was to benefit humanity as a whole, unconstrained by the need to generate financial return; it also said that, as a nonprofit, OpenAI aimed to build value for everyone rather than shareholders and would encourage publication of papers, blog posts, and code while sharing patents with the world (OpenAI²). The same announcement named Sam Altman and Elon Musk as co-chairs and said funders had committed $1 billion, though OpenAI expected to spend only a small fraction in the next few years (OpenAI²).

That language matters because artificial general intelligence, or AGI, was never just another software product in OpenAI’s own framing. OpenAI’s charter later defined AGI as highly autonomous systems that outperform humans at most economically valuable work and said OpenAI’s mission was to ensure that AGI benefits all of humanity (OpenAI Charter³). The charter also committed OpenAI to avoid enabling uses of AI or AGI that harm humanity or unduly concentrate power, while acknowledging that technical leadership was necessary because safety advocacy alone would be insufficient (OpenAI Charter³). That is the founding bargain in plain English: OpenAI would try to be powerful enough to matter, but not captured by the usual incentives of power.

The strongest defense is real. By 2019, OpenAI said the frontier had become far more capital-intensive than expected. Its announcement creating OpenAI LP said the most dramatic AI systems required enormous computational power and that OpenAI would need to invest billions of dollars in cloud compute, talent, and AI supercomputers (OpenAI⁴). The new capped-profit company allowed investors and employees to receive capped returns, sent excess returns to the nonprofit, placed OpenAI LP under nonprofit board control, and said the company’s primary fiduciary obligation was to the OpenAI Charter, even at the expense of financial stakes (OpenAI⁴). First-round investor returns were capped at 100 times investment, which is a very generous cap, but still not ordinary unlimited shareholder primacy (OpenAI⁴).

This is where I part company with the easy anti-profit line. A pure donation-funded lab probably could not have trained and deployed frontier models at the scale OpenAI reached. Microsoft’s first $1 billion partnership with OpenAI in 2019 was explicitly about building Azure supercomputing technologies and making Microsoft OpenAI’s preferred partner for commercializing new AI technologies (Microsoft⁵). In 2023, Microsoft announced a multiyear, multibillion-dollar extension that made Azure OpenAI’s exclusive cloud provider across research, products, and API services (Microsoft⁶). If the only options were nonprofit purity and technical irrelevance, I would choose the messy hybrid.

But those were not the only options. OpenAI could have been blunt from the beginning that the public-benefit project might become a conventional equity story for insiders, investors, and a dominant cloud platform. It did not. The 2019 structure was sold as a way to preserve mission control while raising money. The hard question is whether that control worked when tested.

The 2023 crisis is the key exhibit. In November 2023, OpenAI’s nonprofit board removed Altman as CEO. The board later said Altman had not been consistently candid in his communications, and OpenAI’s 2024 summary of the WilmerHale review said the prior board acted within its broad discretion but moved on an abridged timeline and did not anticipate destabilizing the company (OpenAI⁷). The same review summary said the firing did not arise from product safety, security, development pace, finances, or statements to investors and partners, but from a breakdown in trust between the prior board and Altman (OpenAI⁷).

That finding helps OpenAI in one narrow sense. It weakens the claim that Altman was fired because he was recklessly racing ahead on safety. But it hurts OpenAI in the larger governance story. If the nonprofit board could not remove the CEO over a trust breakdown without nearly losing the institution, how credible was its power to overrule management on a revenue-critical deployment or a Microsoft-sensitive strategic decision?

The employee revolt answered that question. Axios reported that an overwhelming majority of OpenAI’s 700-plus employees signed a letter demanding the board resign and threatening to join a new Microsoft research unit unless the board stepped down (Axios⁸). Bloomberg and other outlets put the number above 700 of roughly 770 employees (Bloomberg via The Star⁹). The board folded, Altman returned, and OpenAI later added new governance procedures, including a strengthened conflict-of-interest policy and a Mission & Strategy committee (OpenAI⁷).

A governance structure that survives only when it does not seriously cross employees, investors, and Microsoft is not mission control. It is mission consent.

The money makes the myth harder to defend. On May 4, 2026, AP reported that Brockman disclosed in court that his OpenAI stake was worth nearly $30 billion and that he said he had not personally invested money in OpenAI; AP also described OpenAI as having evolved from a 2015 nonprofit startup primarily funded by Musk into a venture valued at $852 billion (Associated Press¹⁰). Reuters separately reported that Brockman disclosed additional financial ties to Altman, including stakes in Altman-backed startups and a past stake in Altman’s family office worth $10 million at the time (Reuters via Investing.com¹¹). None of that proves fraud. It does prove that the nonprofit-origin project produced extraordinary private claims on value.

OpenAI’s 2025 restructuring did not solve the moral problem. It clarified it. OpenAI says the nonprofit is now the OpenAI Foundation, the for-profit is OpenAI Group PBC, and the Foundation continues to control the group through special voting and governance rights (OpenAI¹²). The Foundation holds a 26 percent equity stake worth about $130 billion, while Microsoft holds roughly 27 percent and current and former employees and investors hold the remaining 47 percent (OpenAI¹²). The Microsoft-OpenAI agreement also says Microsoft’s stake is worth about $135 billion, preserves Microsoft as OpenAI’s frontier model partner, extends Microsoft’s model and product IP rights through 2032, and includes an incremental $250 billion Azure services commitment from OpenAI (OpenAI¹³).

That is not a nonprofit lab with a commercial subsidiary. It is a commercial AI empire with a mission-holding foundation wired into its cap table.

The best counterargument is that this is what institutional seriousness looks like in frontier AI. Compute is expensive. Talent is scarce. Google, Meta, Anthropic, Microsoft, and state-backed efforts are not waiting politely for a charity to pass the donation hat. If OpenAI had refused commercial capital, the frontier might simply have moved to companies with fewer public-benefit constraints.

I buy the premise and reject the conclusion. The real test of a public-benefit AI institution is not whether it can raise money when mission and growth point in the same direction. The test is whether it can say no when they diverge. OpenAI’s history gives me little reason to believe the nonprofit mission has that kind of operational veto. The board crisis showed that formal authority can buckle under talent and platform pressure. The Microsoft agreements show that OpenAI’s path to scale is deeply entangled with a commercial partner’s infrastructure, IP rights, and strategic interests. The insider wealth shows that capped-profit rhetoric can still yield private fortunes large enough to swallow the word capped.

The jury may still find that Musk did not prove his legal claims. A denial of summary judgment, as reported in the January 2026 order, left disputes for trial rather than deciding that OpenAI betrayed anyone (FindLaw¹⁴). Musk is also an imperfect messenger, given his own AI venture and his long record of trying to control strategic assets. But the public should not let the messenger settle the myth.

My prediction is this: over the next 18 months, OpenAI will point often to Foundation control and public-benefit language, but the meaningful signal will be narrower. Watch for one concrete case in which the Foundation or its Safety and Security Committee blocks, delays, or materially narrows a major model release, Microsoft-linked commercial deal, or revenue plan over management opposition. If that does not happen by April 2027, the founding myth will have been settled outside the courtroom: OpenAI did not preserve the nonprofit bargain. It converted it into a brand of governance that works only when profit agrees.