Lessons from OpenAI: Boards and the Spin of Corporate Governance

Little was known about OpenAI until the launch of ChatGPT, its generative Artificial Intelligence (“AI”) chatbot, in November 2022. It catapulted the organization into prominence and both OpenAI and Sam Altman, then Chief Executive Officer (“CEO”), instantly became household names. Almost a year later, on November 17, 2023, Altman was briefly ousted from OpenAI by its Board of Directors (“Board”). The next few days witnessed over 95% of its employees threatening to leave and demands for Altman’s return to the Board. Altman was ultimately reinstated as the CEO and a new Board was set up at OpenAI, closing the curtains on this saga.

Altman’s brief departure from OpenAI presents lessons in corporate governance, particularly in the context of a board of directors’ powers and responsibilities. In this note the authors analyze OpenAI’s governance structure and revisit the period of Altman’s firing, while briefly reviewing the role of a board of directors and shareholders in considering the removal of a director in an Indian context.

UNDERSTANDING OPENAI

OpenAI was founded as a non-profit in 2015 with the core mission of ensuring that artificial general intelligence (AGI) benefits all of humanity. Succeeding in this mission would require advanced technologies, supercomputing capabilities and highly skilled talent, which demanded billions in investment. Since the non-profit structure did not permit raising capital, OpenAI created OpenAI LP (“First Restructuring”) as a hybrid of

1 Since March 11, 2019, colloquially “OpenAI” refers to OpenAI LP, and the original entity is referred to as “OpenAI Nonprofit.” For the purposes of this note, OpenAI refers to the entire organization and the authors have sought to provide clarifications wherever necessary.

2 These engagements with OpenAI stopped after Elon Musk sued the organization, after their move towards a for-profit structure, alleging, among other things, racketeering activities and for abandoning their original mission statement and duping investors.

3 The law firm was retained on December 8, 2023 and completed their review three months later on March 8, 2024. a for-profit and non-profit, a ‘capped-profit’ company where returns to investors and employees would be capped. OpenAI LP’s operating agreement captures this approach, empowering the company to re-invest all cash flow into research and development, and states that the organization “…may never make a profit…and is under no obligation to do so.”

Corporate Structure

OpenAI, L.P. will be a for-profit Delaware Limited Partnership managed by its General Partner, a single-member Delaware LLC controlled by OpenAI, Inc. (the Nonprofit)’s Board of Directors. At all times, no more than a minority of the Nonprofit’s Board of Directors will be holders of any economic interest in OpenAI, L.P. or in the employee holdings entity (described below).

Source: OpenAI website

This was done to incentivize research, development and deployment of AGI in alignment with OpenAI’s mission objectives while balancing commerciality, safety and sustainability. Employees working on for-profit initiatives, almost the entirety of OpenAI’s employee pool, were transitioned over to OpenAI LP. 1 The governing structure of the for- profit involves a manager entity (OpenAI GP LLC) which is wholly owned and controlled by the non-profit, which in turn is governed by the Board and its majority independent directors who do not hold equity in OpenAI. Effectively, the non-profit, through its Board, would govern and oversee all activities at OpenAI.

The First Restructuring immediately eased OpenAI’s cash crunch, following a USD 1 billion investment from Microsoft in July 2019. Over the years, regular investments helped the organization continue its work, while simultaneously increasing its valuation. Recently, the organization secured USD 6.6 billion in funding at a post-money valuation of USD 157 billion. Reports of another restructuring, resulting in a for-profit OpenAI not controlled by the Board (“Second Restructuring”), coupled with significant financing arrangements, have raised concerns among early investors, donors and policy experts regarding the organization’s commitment to its mission statement.

THE BOARD OF DIRECTORS

The original Board in 2015, during OpenAI’s inception, included, among others, Sam Altman and Elon Musk. By 2017, Greg Brockman and Ilya Sutskever (co-founders of OpenAI) had joined the Board, along with Holden Karnofsky (Executive Director of Open Philanthropy), following a staggered grant of USD 30 million towards OpenAI by Karnofsky’s foundation. Musk left the Board in 2018, reportedly due to a potential conflict, but was said to continue donating to and advising OpenAI. 2

When the First Restructuring was announced, the Board consisted of Greg Brockman (Chairman & Chief Technology Officer (CTO)), Ilya Sutskever (Chief Scientist), Sam Altman (CEO) and five non-employee independent directors. By the end of 2021, after several changes, Congressman Will Hurd and Helen Toner, Director of Strategy at Georgetown’s Center for Security and Emerging Technology, had joined the Board.

Toner had previously worked with Open Philanthropy and perhaps acted as Karnofsky’s replacement, who had earlier departed from the Board citing a potential conflict (his wife and brother-in-law were deeply involved in starting Anthropic, a direct competitor to OpenAI). Subsequently, Hurd resigned to pursue his political career. On November 17, 2023, the Board consisted of six members – Adam D’Angelo, Helen Toner, Tasha McCauley, Ilya Sutskever (Chief Scientist), Greg Brockman (Chairman) and Sam Altman (CEO). Later that day, Altman and Brockman were fired, leaving four Board members. Altman was fired on the basis that he had lost confidence of the Board, following a deliberative review process, “…which concluded that he was not consistently candid in his communications with the board, hindering its ability to exercise its responsibilities.” Mira Murati, then CTO, was appointed interim CEO and search for a permanent successor was underway. The Board was tasked to oversee all OpenAI activities, ensuring that the organization’s mission was not compromised. Its self- governing structure demanded that all decision makers be accountable to the majority independent Board.

SAM ALTMAN’S RETURN AND THE BOARD OVERHAUL

The aftermath following Altman’s departure was eventful for OpenAI. Satya Nadella, CEO of Microsoft, offered to engage Altman’s (and Brockman’s) expertise and services to build and research AI products as part of Microsoft. Additionally, almost all OpenAI employees, including senior management, publicly stated their support for Altman— demanding his return, else they would follow him to Microsoft’s new venture. In the larger interest of protecting the organization, the Board reversed their decision and Altman returned as CEO.

The entire Board, barring D’Angelo, would be overhauled with Altman’s return. A new, diverse and qualified Board was to be built “whose collective experience represents the breadth of OpenAI’s mission – from technology to safety to policy,” that would include a non-voting observer for Microsoft (“New Board”). An independent committee of the New Board was set up to oversee a review of the events leading to Altman’s removal, conducted by a leading law firm. 3

The review included interviews with prior Board members, OpenAI executives and witnesses, analysis of over 30,000 documents, and evaluation of various corporate actions. Findings suggested that the Board’s decision “did not arise out of concerns regarding product safety or security, the pace of development, OpenAI’s finances, or its statements to investors, customers, or business partners.” Rather, it was the breakdown in trust between the prior Board and Altman that precipitated his firing. It was the Board’s belief at the time that this would mitigate internal management challenges, but they did not anticipate the potential destabilization of the Company.

Interestingly, the review did not dispute any of the reported claims surrounding accountability and oversight. However, it concluded that “Altman’s conduct did not mandate removal.” The basis for this conclusion seems unclear since Altman was fired because of “loss of confidence” and “breakdown of trust” and not because of any alleged misconduct. The New Board endorsed Altman’s return and announced adoption of improved corporate governance guidelines focused on implementation andadvancement of OpenAI’s core mission.

REVIEWING DIRECTORS’ REMOVAL FROM THE INDIA LENS

Several companies in India have, much like Altman, larger-than-life directors at the helm. A large majority of these are historically family-run businesses with a promoter- controlled board of directors and, more importantly, majority shareholding / controlling stake in the company. Although India has witnessed director removals, unlike OpenAI, the board of directors does not play the pivotal role. Subject to their articles of association, removal of a director from an Indian company requires shareholders’ approval, by an ordinary resolution (>50%) vote and cannot solely be decided by a company’s board of directors.

The procedure for director removal in India involves approvals at the board of directors’ and shareholders’ level. A board resolution must be passed,
(i) approving removal of such director; and
(ii) calling for a shareholders’ meeting for approval by shareholders pursuant to an ordinary resolution. Indian company law also requires that the director sought to be removed be given an opportunity of being heard (and make a representation) before the company considers the resolution for approval. However, there is no requirement under Indian company law for reasons to be given for removal of a director; such removal will be upheld if due process has been followed and the shareholders have approved such removal by the requisite majority. Once the shareholders’ resolution is approved, the company must complete statutory filings and update their registers to record the relevant changes. Alternatively, in terms of their articles of association, a company’s board of directors may be empowered to remove a director in accordance with due procedure.

CONCLUSION

OpenAI’s corporate structure emphasized non-profit humanitarian goals, with a majority independent self-governing board of directors. In theory, this structure presents a powerful governance model where checks and balances can be placed on sensitive decision making. A loss of confidence in a director, due to lack of transparency and accountability, would perhaps give rise to circumstances where a board of directors’ discretionary removal powers may be used. However, the futility of Altman’s removal highlights a failure of such self-governance models in today’s corporate reality.