In late November 2023, OpenAI, the non-profit artificial intelligence research company, found itself embroiled in a leadership crisis that sent shockwaves through the AI community. The sudden removal of CEO Sam Altman by the company's board of directors triggered a backlash from employees, investors, and prominent AI researchers, leading to his reinstatement just days later. The incident exposed deep-seated tensions within OpenAI regarding its mission, governance, and the role of its influential CEO. This essay delves into the recent turmoil at OpenAI, examining the corporate governance issues that surfaced and the lessons learned for the future of the organization.
On November 17, 2023, OpenAI's board of directors made the unexpected decision to oust Altman, the company's co-founder and CEO since 2015. The decision sent shockwaves through OpenAI, with hundreds of employees reportedly threatening to resign in solidarity with Altman. Several prominent AI researchers also expressed their support for Altman, questioning the board's judgment and the future of OpenAI. Amidst mounting pressure, the board reversed its decision just days later, reinstating Altman as CEO. The board acknowledged that its communication about the removal had been poorly handled and emphasized its commitment to Altman's leadership.
The leadership crisis at OpenAI highlighted tensions and governance issues that had been simmering beneath the surface. One key issue was the balance between OpenAI's mission as a non-profit research organization and its desire to further advance its technology.
Many speculations arose as to why Altman was ousted in the first place. Since the Board of Directors were majority independent and did not hold equity in OpenAI, it was said that the tension occurred mainly due to disagreements over the direction of Artificial General Intelligence (AGI). Altman is a big proponent developing AGI at full throttle, whereas Ilya Sutskever and the other board members preferred a more cautious speed treading the AI water. The New York Times also recounted a leadership argument between Altman and Helen Toner over an academic paper authored by her, one of the former independent.
The fact that this disagreement can lead to the firing of a CEO exposed interesting features of OpenAI corporate governance and the distribution of control. As shown in the diagram, a regular for-profit corporation has three main layers: shareholders (investors who purchased the shares of the company but not involved in the day-to-day operations), the board of directors (overseeing the management and business strategies), and management (led by the CEO, executing the strategies and the everyday operations of the company). As a non-profit organization, OpenAI’s governance structure is quite unique. OpenAI Inc., the nonprofit, is controlled by the Board of Directors, which was consisted of mainly independent directors. OpenAI, Inc., along with employees and investors, owns shares in the holding company, which owns 51% of the for-profit entity, OpenAI Global, but do not possess voting power. Hence, the control over OpenAI is theoretically, completely, in the hands of the Board.
The sudden removal and reinstatement of Altman raised questions about the effectiveness of OpenAI's governance structure. The board's decision-making process, lack of transparency and communication during the crisis damaged its credibility and raised concerns about its ability to oversee the company effectively.
Too much control in the hands of the Independent Board? The abrupt ousting of Altman displayed the overwhelming control held by the independent board members who had no apparent monetary stake in the development of OpenAI. One can argue that the motivations of people who sit on a nonprofit’s Board are generally benevolent and in the interest of the nonprofit’s mission. Similarly, at OpenAI, the Board’s main fiduciary duty is to create safe AGI. Nonetheless, that does not guarantee that the decisions made by the Board will benefit the organizations’ development. Furthermore, the current board structure was not in its originally proposed form, but rather a remainder after the departure of three members, Reid Hoffman, Shivon Zilis, and Will Hurd. The reduced number of Board members largely shifted the power dynamic and further destabilized the Board.
The types of stakeholders involved also became incredibly complicated after OpenAI morphed into its current structure, with a governing nonprofit, a holding company, and a for-profit sub-entity that ties closely with Microsoft. The incentives of Altman, Microsoft, investors in the holding company, and the employees, and the (original) Board before the ouster very likely misaligned. In this situation, it would be in everyone’s interest to have a Board that include members from each side to sit at the table, but the reality was that the Board overpowered everyone else in control.
Did the Board think this through? The answer is clearly no. When the news of Altman’s ouster first broke, many conspiracy theories appeared. However, the messy ways that the Board dealt with the reactions from Greg Brockman, the employees, subsequent meetings with Altman indicated that they were not prepared for a scenario where Altman is not the CEO and this decision was made in a haste. No communication (investor relations and public relations) plans were in place. At the time of Altman’s exit, the plan was to have the CTO, Mira Murati, as its interim CEO, but three days later, the Board offered the job to Twitch co-founder Emmett Shear instead. This quick shift across the span of a few days painted a clear picture that the Board did not know who they wanted and how they wanted to proceed; rather the prominent goal at the time was to remove Altman from a decision-making seat. It also revealed the lack of checks and balances in OpenAI’s leadership structure. In the event of a profound disagreement, no strategy was developed to effectively resolve the disagreement.
Is the new Board for the better? Of the original Board, Altman returned to the company but did not reinstate as a Board member, Sutskever, Toner and McCauley left, and only Adam D’Angelo stayed. New members included Bret Taylor, a former executive at Facebook and Salesforce, and Lawrence H. Summers, a leading economist who held critical positions in the Clinton administration and at Harvard. It is said that D’Angelo and Taylor are critical figures at negotiation tables, and Summers may be a critical proponent who stands on Altman’s side. The direction OpenAI and AGI will go from now on may be changed due to this shift in leadership, but the structure has not changed. Should disagreements arise again, whether similar shakeups happen remains to be seen.
We as investors hate to see a venture going down the garbage chute due to internal disputes. Future founders in the for-profit or non-profit world need a solid corporate governance structure and a clear conflict resolution strategy. For our entrepreneurial and investor community, we would like to suggest the following:
· Establish a reliable governance structure: Ensuring that the incentives align for those who are in control over the organization is critical. no matter what the corporate structure is, all stakeholders that have (quite literally) a stake in the organization should have some form of control or vote over important issues around the organization’s development. In this case, the C-suite, investors, partners, the employees are all affected by the peaks and troughs in OpenAI’s development. Additionally, while we disagree with how the original Board handled the ousting, we still think it is crucial to have a variety of viewpoints especially around an impactful topic, AGI.
· Form an effective conflict resolution strategy: when checks-and-balances are in place, disagreements and conflicts are bound to occur. This is exactly why a potent conflict resolution strategy is necessary. The goal of the strategy is not to get rid of one side and keep the other, but to capitalize on the merits from both (or multiple) sides. An important element of conflict resolution is to minimize hurt feelings. Ultimately, we are all human, and humans have emotions. Negative emotions can get in the way of successful conflict resolution and may exacerbate the fissure. Therefore, an adequate conflict resolution strategy can safeguard the growth of the organization in a positive direction.
· Construct a clear contingency and succession plan: Things change, and people change – many factors can lead to the removal of a CEO. It is important to develop have a clear contingency plan in advance that maintains the proper daily operation of the organization with the sudden removal of executives. This contingency plan should include the change in communication channels that would originally go through the executives, new operating and managing models, and a succession plan. The succession plan, in turn, should include the interim candidates for the executive positions, as well as the selection process, timeline, and criteria for the new executive. This plan should ensure the more efficient selection of new executives without compromising the necessary due diligence of the candidates’ background, integrity, and fit for the position.
The recent turmoil at OpenAI highlighted the challenges and complexities of governing a complex organization such as OpenAI. The company must navigate the delicate balance between pursuing its mission, ensuring financial sustainability, and maintaining internal cohesion. The story of OpenAI also highlighted the importance of a reliable governance, an effective conflict resolution strategy and a clear contingency and succession plan.
.png){kind=icon}
