Corporate Governance

The Public Benefit Corporation: OpenAI, Mission Drift, and the Limits of Governance by Design

OpenAI's transformation from a nonprofit research lab to a $500 billion public benefit corporation is the most consequential corporate restructuring in AI history — and a test of whether governance by design can survive the pressures it was designed to contain.

P J Laszkowicz

On October 28, 2025, OpenAI announced that it had completed the most significant corporate restructuring in the history of artificial intelligence.1 The company that began as a nonprofit research laboratory dedicated to ensuring that artificial general intelligence "benefits all of humanity" was now a for-profit public benefit corporation valued at approximately $500 billion.2 The nonprofit that founded it — renamed the OpenAI Foundation — held 26 percent of the new entity, worth roughly $130 billion, making it one of the best-resourced philanthropic organisations in the world.3

The restructuring had taken years, survived a boardroom coup, weathered regulatory scrutiny from two state attorneys general, and prompted one of the most consequential debates in technology governance: what happens when the institution built to steward the most powerful technology in human history decides it needs to become a conventional corporation to do so?

This article is not about whether OpenAI made the right decision. It is about what the decision reveals: the structural tension between mission-driven governance and the capital demands of frontier AI development, and the question of whether any corporate form — however carefully designed — can hold that tension in place when billions of dollars and geopolitical ambitions are at stake.

From research lab to $500 billion company

The trajectory from nonprofit research lab to half-trillion-dollar corporation took exactly ten years. Each structural transformation was a response to the same underlying problem: the mission required resources that the current structure could not attract.

OpenAI was founded in December 2015 as a nonprofit research organisation.4 Its co-founders included Sam Altman, then president of the startup accelerator Y Combinator, and Elon Musk, along with Ilya Sutskever, Greg Brockman, and several other researchers and entrepreneurs.5 Donors including Altman, Musk, Reid Hoffman, Peter Thiel, and Amazon Web Services pledged $1 billion in capital.6 The stated mission was to develop artificial general intelligence safely and to ensure its benefits were widely distributed. The nonprofit structure was the point: no investor returns, no shareholder pressure, no fiduciary duty to anyone except the mission itself.

The first structural crisis arrived within four years. By 2018, it had become clear that training competitive AI models required computational resources that a nonprofit could not sustain. In March 2019, OpenAI created a "capped-profit" subsidiary, OpenAI LP, controlled by the nonprofit parent, OpenAI Inc.7 Investors could earn returns, but those returns were capped at 100 times their investment — a figure that seemed generous until the company's valuation made the cap irrelevant.8 The structure preserved the nonprofit board's ultimate authority: OpenAI LP's governing documents stated that if the board ever concluded that the organisation's activities were inconsistent with its mission, it could redirect the company's course, up to and including dissolving it entirely.

Microsoft entered the picture in July 2019 with a $1 billion investment, becoming OpenAI's exclusive cloud computing partner through Azure.9 In January 2023, Microsoft committed a further $10 billion in what it described as the "third phase" of the partnership.10 By that point, Microsoft had invested approximately $13.8 billion in total.11 The arrangement was unusual: Microsoft received exclusive API distribution rights and cloud computing obligations, but no seat on the nonprofit board and no formal governance role.

Then, on November 17, 2023, the structure was tested — and it broke.

The five days

OpenAI's board of directors — comprising chief scientist Ilya Sutskever, Quora CEO Adam D'Angelo, entrepreneur Tasha McCauley, and Georgetown CSET researcher Helen Toner — removed Sam Altman as CEO.12 The board's terse statement said only that Altman "was not consistently candid in his communications with the board."13

What followed was five days of corporate chaos that laid bare every tension the nonprofit structure was supposed to manage. Microsoft, which had invested billions but had no board representation, was reportedly informed minutes before the announcement.14 More than 700 of OpenAI's 770 employees signed a letter threatening to resign and join Microsoft if Altman was not reinstated.15 Sutskever publicly reversed himself on November 20, saying he "deeply regrets" his role in the firing.16

On November 21, OpenAI announced a deal for Altman's return. The price was a reconstructed board: Bret Taylor, former Salesforce co-CEO, as chair; former Treasury Secretary Larry Summers; and D'Angelo as the sole holdover.17 McCauley, Toner, and Sutskever were gone. The board that had exercised its governance authority — the mechanism that the nonprofit structure was specifically designed to enable — was itself removed within days.

Sutskever, the chief architect of the firing and one of the company's most important researchers, quietly departed in May 2024.18

The lesson was unambiguous. The nonprofit board had the formal authority to remove the CEO. It exercised that authority. The result was not a governance correction but a governance collapse. The board that prioritised safety concerns over commercial momentum was replaced by a board that would not repeat the experiment.

The restructuring

What followed was a two-year negotiation over what OpenAI should become. In late 2024, the company proposed converting to a standard for-profit corporation, eliminating the nonprofit's controlling role entirely.19 The backlash was immediate. Elon Musk filed a lawsuit. The attorneys general of California and Delaware — the states where OpenAI was incorporated and headquartered — signalled that they would scrutinise any conversion that stripped the nonprofit of its assets and mission.20 A coalition called "Not for Private Gain" organised opposition from governance scholars, former employees, and civil society groups.

On May 5, 2025, OpenAI pivoted. The for-profit entity would become a Delaware public benefit corporation rather than a standard C-corporation, and the nonprofit would retain control.21 The restructuring was completed on October 28, 2025, after both attorneys general signed memoranda of understanding with the company.22

The architecture of compromise

The resulting structure is a nested set of compromises, each responding to a different constituency.

The OpenAI Foundation — the renamed nonprofit — holds 26 percent of OpenAI Group PBC, the for-profit entity.23 That stake is worth approximately $130 billion.24 The Foundation's governance powers are, on paper, extensive: it holds the sole authority to appoint and remove all members of the PBC's board of directors.25 The Safety and Security Committee, which has the authority to "require mitigation measures — up to and including halting the release of models or AI systems — even where the applicable risk thresholds would otherwise permit release," sits under the Foundation, not the PBC.26

Microsoft holds 27 percent of the PBC, valued at roughly $135 billion.27 It retains exclusive API distribution rights through Azure and remains OpenAI's primary cloud computing partner. But the partnership terms shifted: OpenAI is no longer bound by a Microsoft right of first refusal on compute, and has contracted to purchase $250 billion in Azure services.28 Microsoft holds no governance power — no board seats, no veto rights, no formal role in safety or mission oversight.

The remaining 47 percent is held by employees and other investors.29

The attorneys general extracted twenty concessions before approving the restructuring.30 Both the Foundation and the PBC must remain headquartered in California. The PBC's stated mission must be identical to the Foundation's. The Safety and Security Committee must be empowered to halt product releases. And both states promised ongoing monitoring.31

The structure looks robust on paper. Whether it is robust in practice depends on questions that corporate law cannot answer.

What a public benefit corporation is (and isn't)

A Delaware public benefit corporation is a relatively recent legal innovation. Under Delaware's General Corporation Law, Subchapter XV, a PBC must identify a specific public benefit in its charter and its directors must balance three interests: the pecuniary interests of stockholders, the interests of those materially affected by the company's conduct, and the company's identified public benefit.32

This is a meaningful departure from the standard corporate form, in which directors' fiduciary duties run primarily to stockholders and are assessed through the lens of shareholder value. In a PBC, directors who make decisions that sacrifice some shareholder returns in favour of the stated public benefit are, in principle, legally protected from the shareholder lawsuits that would otherwise follow.33

The protection is real but untested. As of late 2025, no shareholder has sued over the management of a PBC's competing obligations.34 There is no case law on how courts should weigh pecuniary interests against public benefit when the two conflict. The balancing standard — that a director's decision must be "informed and disinterested and not such that no person of ordinary, sound judgment would approve" — provides wide discretion.35 That discretion is an asset when directors are acting in good faith. It is a vulnerability when they are not.

The PBC form has been adopted by a number of notable companies. Patagonia's 2022 conversion, in which founder Yvon Chouinard transferred ownership to a purpose trust and a nonprofit, gave the model significant credibility as a tool for mission protection at scale.36 Kickstarter reincorporated as a PBC in 2015. Lemonade, the insurance company, went public as a PBC in 2020. But none of these companies approached OpenAI's $500 billion valuation, none operated in an industry with the geopolitical significance of frontier AI, and none faced the specific tension between safety imperatives and commercial pressure that defines AI development.

OpenAI Group PBC is, by a very large margin, the most valuable public benefit corporation ever created. It is an experiment in corporate governance with no precedent and no fallback.

The overlap problem

The structural criticism of OpenAI's restructuring is not that the wrong form was chosen, but that the form's safeguards are undermined by the people who occupy its positions.

Nearly all board members serve simultaneously on both the Foundation and the PBC boards.37 The Foundation is supposed to oversee the PBC. The PBC is supposed to be overseen by the Foundation. When the same individuals sit on both sides of this relationship, the oversight function collapses into self-supervision.

Alnoor Ebrahim, a nonprofit governance scholar at Tufts University, described this as a fundamental governance failure.38 The Foundation holds 26 percent equity — substantial in dollar terms, but representing only a quarter of the entity it nominally controls. Ebrahim noted that this stands in unfavourable comparison to the Health Net precedent, in which 80 percent of equity was transferred to a health foundation during a nonprofit-to-for-profit conversion, or The Philadelphia Inquirer model, in which a nonprofit maintained majority control.39

Judith Bell, an impact officer at the San Francisco Foundation, was more direct: "There's a bazillion conflicts of interest here."40

Steven Adler, a former leader on OpenAI's safety team, expressed doubt about the Safety and Security Committee's practical independence. The committee has the formal authority to halt model releases. But formal authority and operational independence are different things. "I hope that a truly independent body will do a better job," Adler said, while acknowledging uncertainty about whether committee members could maintain autonomy under organisational pressures toward profitability.41

The Eyes On OpenAI coalition, led by director Orson Aguilar, was more blunt: "The nonprofit continues to operate under the influence of the for-profit it supposedly oversees."42

The mission statement

Between 2023 and 2024, OpenAI quietly deleted the word "safely" from its mission statement.43

The previous formulation read: "to build general-purpose artificial intelligence that safely benefits humanity, unconstrained by a need to generate financial return."44 The current version reads: "to ensure that artificial general intelligence benefits all of humanity."45

The deletion of a single word might seem minor. It is not. The word "safely" was the linguistic anchor that connected the mission statement to the specific concern — the risk of advanced AI systems — that justified the nonprofit structure in the first place. Its removal was noted by Ebrahim, who described it as signalling a shift in priorities from safety toward commercial objectives.46

OpenAI's leadership has not publicly explained the change. The company's blog posts and structural documents from 2025 use the shorter formulation without comment. The Safety and Security Committee's formal powers remain in place. But the governing documents of both the Foundation and the PBC now describe a mission that could, linguistically, be satisfied by beneficial AI that is not safe — or by safe AI whose benefits accrue to investors rather than humanity broadly.

Words in mission statements are not decorative. They are the standard against which fiduciary duties are measured, board decisions are justified, and attorneys general exercise oversight. When the standard changes, the accountability framework changes with it.

The safety committee question

The Safety and Security Committee is the structural element most often cited as evidence that the restructuring preserves meaningful safety governance. Its formal powers are significant: it sits under the Foundation (not the PBC), it is chaired by Dr. Zico Kolter (a machine learning researcher at Carnegie Mellon), and its mandate includes the authority to require mitigation measures up to and including halting model releases, even when the applicable risk thresholds would otherwise permit them.47

On paper, this is the strongest safety governance mechanism at any major AI company. No other frontier AI lab has a committee with the formal power to unilaterally halt a product release, housed in a separate legal entity from the commercial operation, and overseen by an independent researcher.

The question is whether formal power translates to operational reality.

The committee's effectiveness depends on three conditions. First, it must have timely access to the technical information needed to assess risk — not after a model is ready for release, but during development. Second, its members must be willing to exercise their authority against the commercial interests of a $500 billion company whose employees hold 47 percent of its equity. Third, the Foundation board must be willing to defend the committee's decisions if they are contested.

None of these conditions are guaranteed by the corporate structure. They depend on norms, culture, and the personal resolve of individuals — precisely the factors that failed during the November 2023 crisis.

The board that fired Altman had formal authority. It exercised that authority. It was replaced within days by a board that would not. The Safety and Security Committee has formal authority that exceeds what the old board had. Whether it would survive exercising that authority against the will of management, investors, and a workforce with billions in equity is an open question.

Governance by design, tested by reality

Four months after the restructuring was completed, the new structure faced its first test.

On February 27, 2026, the United States government designated Anthropic — OpenAI's principal competitor in frontier AI — a supply chain risk and terminated its military contracts over Anthropic's refusal to remove safety restrictions on autonomous weapons and mass surveillance.48 Within hours, OpenAI announced that it had secured the Pentagon contract.49

Sam Altman moved quickly to manage the optics. In a memo to staff, he wrote that OpenAI would draw the same red lines that Anthropic had defended: no AI for mass surveillance of American citizens, no fully autonomous lethal weapons.50 On February 27, the Pentagon agreed to OpenAI's terms, and the contract was amended to include language stipulating that "the AI system shall not be intentionally used for domestic surveillance of U.S. persons and nationals."51

But by March 3, Altman himself acknowledged the damage. "We were genuinely trying to de-escalate things and avoid a much worse outcome," he wrote, "but I think it just looked opportunistic and sloppy."52 He admitted he "shouldn't have rushed" to announce the deal on the same day Anthropic was sanctioned.53

The episode revealed something important about the PBC structure's limits. Altman's red lines — articulated in a staff memo and a social media post — are substantively identical to the contractual terms that Anthropic was punished for maintaining. But Anthropic's terms were binding contractual provisions, enforceable under the original agreement. Altman's are public commitments, voluntarily adopted and voluntarily amendable. The amended contract language covers domestic surveillance of US persons, but the original Anthropic restrictions were broader, covering mass surveillance and autonomous weapons targeting without geographical limitation.54

The PBC structure did not prevent OpenAI from taking the contract. The Safety and Security Committee did not publicly weigh in. The Foundation board did not intervene. The governance architecture — 20 concessions from two attorneys general, a mission-aligned PBC charter, an empowered safety committee — was not visibly activated.

This is not necessarily a failure of the structure. It is possible that the committee reviewed the contract and found it acceptable. It is possible that the Foundation board considered the matter and concluded that a Pentagon contract with safety restrictions served the public benefit. But the absence of any visible governance process is itself informative. When the first significant test arrived, the public-facing response came from the CEO's social media account, not from the governance mechanisms designed for exactly this purpose.

The test that keeps arriving

The OpenAI restructuring is the most ambitious attempt to embed mission-driven governance into a commercial AI company. It is more sophisticated than any previous nonprofit-to-for-profit conversion. The Foundation's formal powers — board appointment, safety committee authority, mission alignment requirements — exceed what any other AI company's governance structure provides.

But formal powers are scaffolding, not architecture. The November 2023 crisis demonstrated that a board with the authority to remove a CEO could itself be removed when it exercised that authority. The February 2026 Pentagon episode demonstrated that a safety committee with the authority to halt model releases did not visibly engage when the company's most consequential policy decision was made in a matter of hours.

The structural critics are not wrong to point to the overlap problem, the mission statement change, and the enforcement gaps. The structural defenders are not wrong to note that the Foundation's powers are real, the Safety and Security Committee's mandate is the strongest in the industry, and the attorneys general are actively monitoring compliance.

Both observations can be true simultaneously. The structure is the best available. It may also be insufficient.

The question that the OpenAI restructuring poses is not whether a public benefit corporation is better than a standard C-corporation for governing frontier AI. It plainly is. The question is whether any corporate form — however carefully designed, however many concessions are extracted by regulators, however empowered the safety mechanisms — can hold the tension between a $500 billion valuation and a mission that may require saying no to the most powerful customer in the world.

The November 2023 board had the power and the conviction. It lasted five days. The Safety and Security Committee has the power. Whether it will ever need the conviction — and what will happen if it does — is the question that will define whether OpenAI's governance experiment succeeds or becomes the most expensive illustration of mission drift in corporate history.

Footnotes

  1. CNBC, "OpenAI completes restructure, solidifying Microsoft as a major shareholder," 28 October 2025. ↩

  2. OpenAI, "Built to benefit everyone," 28 October 2025. ↩

  3. NBC News, "A nonprofit on top, billions below: How OpenAI's new structure works," 28 October 2025. ↩

  4. OpenAI, "Introducing OpenAI," 11 December 2015. ↩

  5. Additional co-founders included Trevor Blackwell, Vicki Cheung, Andrej Karpathy, Durk Kingma, John Schulman, Pamela Vagata, and Wojciech Zaremba. ↩

  6. Donors pledging the initial $1 billion included Altman, Musk, Greg Brockman, Reid Hoffman, Jessica Livingston, Peter Thiel, Amazon Web Services, and Infosys. ↩

  7. OpenAI, "OpenAI LP," 11 March 2019. ↩

  8. The 100x cap applied to the initial investors in the capped-profit entity. Later investment rounds negotiated different return terms. ↩

  9. Microsoft, "Microsoft invests in and partners with OpenAI to support us building beneficial AGI," 22 July 2019. ↩

  10. CNBC, "Microsoft to invest $10 billion in ChatGPT creator OpenAI, report says," 10 January 2023. ↩

  11. Microsoft's cumulative investment of approximately $13.8 billion included the 2019 investment, the 2023 multibillion-dollar commitment, and interim funding rounds. ↩

  12. The board at the time of Altman's removal comprised Ilya Sutskever (chief scientist), Adam D'Angelo (Quora CEO), Tasha McCauley, and Helen Toner (Georgetown CSET). ↩

  13. OpenAI, board statement, 17 November 2023. ↩

  14. Axios, "OpenAI chaos: A timeline of Sam Altman's firing and return," 22 November 2023. ↩

  15. ABC News, "4 days from fired to re-hired: A timeline of Sam Altman's ouster from OpenAI," 22 November 2023. ↩

  16. Ilya Sutskever, post on X (formerly Twitter), 20 November 2023. ↩

  17. PBS News, "Sam Altman reinstated as OpenAI CEO with new board replacing the one which fired him," 22 November 2023. ↩

  18. Sutskever announced his departure from OpenAI on 14 May 2024. He subsequently co-founded Safe Superintelligence Inc. ↩

  19. ProMarket, "OpenAI Abandons Move to For-Profit Status After Backlash. Now What?" 6 May 2025. ↩

  20. The California attorney general's scrutiny was driven by the state's authority over charitable assets under the Supervision of Trustees and Fundraisers for Charitable Purposes Act. ↩

  21. OpenAI, "Why OpenAI's structure must evolve to advance our mission," 5 May 2025. ↩

  22. California AG MOU signed 27 October 2025; Delaware AG Statement of No Objection issued 28 October 2025. State of Delaware, "AG Jennings completes review of OpenAI recapitalization," 28 October 2025. ↩

  23. EA Forum, "The OpenAI Governance Transition: The History, What It Is, and What It Means," October 2025. ↩

  24. OpenAI, "Built to benefit everyone," 28 October 2025. ↩

  25. The Foundation's sole authority to appoint and remove PBC board members was confirmed in the memoranda of understanding with both attorneys general. ↩

  26. OpenAI, "Our structure," accessed October 2025. The Safety and Security Committee sits under the OpenAI Foundation, not the PBC. ↩

  27. Microsoft, "The next chapter of the Microsoft–OpenAI partnership," 28 October 2025. Microsoft's 27 percent stake is on an as-converted, diluted basis. ↩

  28. OpenAI contracted to purchase an incremental $250 billion in Azure services. Microsoft no longer holds a right of first refusal to be OpenAI's compute provider. Microsoft, "The next chapter of the Microsoft–OpenAI partnership," 28 October 2025. ↩

  29. EA Forum, "The OpenAI Governance Transition," October 2025. ↩

  30. CalMatters, "OpenAI's restructuring deal with California is full of holes, critics say," October 2025. ↩

  31. Both attorneys general promised ongoing monitoring as part of their non-objection to the restructuring. ↩

  32. Delaware General Corporation Law, Title 8, Chapter 1, Subchapter XV, §365. ↩

  33. Under §365, a PBC director is deemed to satisfy fiduciary duties if the director's decision is both informed and disinterested and not such that no person of ordinary, sound judgment would approve. ↩

  34. As noted by the American Bar Association and multiple Delaware corporate law practitioners, no PBC shareholder lawsuit has tested the balancing standard as of late 2025. ↩

  35. Delaware Code, Title 8, §365. ↩

  36. Patagonia transferred 100 percent of its voting stock to the Patagonia Purpose Trust and 100 percent of its nonvoting stock to the Holdfast Collective, a 501(c)(4), in September 2022. ↩

  37. The Conversation, "OpenAI has deleted the word 'safely' from its mission — and its new structure is a test for whether AI serves society or shareholders," 2025. Author: Alnoor Ebrahim, Tufts University. ↩

  38. Alnoor Ebrahim, nonprofit governance scholar at Tufts University, quoted in The Conversation, 2025. ↩

  39. Ebrahim cited Health Net (80 percent equity to health foundation) and The Philadelphia Inquirer (nonprofit majority control) as precedents that transferred greater value and control to mission-driven entities. ↩

  40. Judith Bell, impact officer at the San Francisco Foundation, quoted in CalMatters, October 2025. ↩

  41. Steven Adler, former OpenAI safety team leader, quoted in CalMatters, October 2025. ↩

  42. Orson Aguilar, director of Eyes On OpenAI coalition, quoted in CalMatters, October 2025. ↩

  43. The Conversation, Ebrahim, 2025. ↩

  44. The previous formulation appeared in OpenAI's founding documents and early corporate communications. ↩

  45. OpenAI's current mission statement as published on openai.com and in the PBC charter documents. ↩

  46. Ebrahim, The Conversation, 2025. ↩

  47. The Safety and Security Committee is chaired by Dr. Zico Kolter, a machine learning researcher at Carnegie Mellon University, who also serves on the Foundation board. OpenAI, "Our structure," accessed October 2025. ↩

  48. FASCSA supply chain risk designation, 27 February 2026. ↩

  49. CNBC, "OpenAI strikes deal with Pentagon, hours after rival Anthropic was blacklisted by Trump," 27 February 2026. ↩

  50. Axios, "Sam Altman says OpenAI shares Anthropic's red lines in Pentagon fight," 27 February 2026. ↩

  51. CNBC, "OpenAI's Altman admits defense deal 'looked opportunistic and sloppy' amid backlash," 3 March 2026. The amended language covered domestic surveillance of US persons and nationals. ↩

  52. Altman, post on X, 3 March 2026, quoted in CNBC. ↩

  53. Altman stated he "shouldn't have rushed" to announce the deal on the same day. CNBC, 3 March 2026. ↩

  54. Anthropic's Acceptable Use Policy prohibited mass surveillance and fully autonomous weapons targeting without restriction to US persons or domestic contexts. NPR, "OpenAI says it shares Anthropic's 'red lines' over military AI use," 27 February 2026. ↩