OpenAI Board’s Search for New Directors Includes a Rival
Last month, OpenAI board director Adam D’Angelo called Databricks CEO Ali Ghodsi to see if he might consider joining OpenAI’s board, according to a person familiar with the situation. That kind of request might not seem unusual—except that Ghodsi has positioned Databricks as a kind of anti-OpenAI through its business of helping companies develop AI applications instead of relying on OpenAI’s technology.
Ordinarily, companies don’t invite representatives of quasi rivals to join their board. Even if Ghodsi doesn’t end up joining, the fact that D’Angelo made such an approach signals the potential for further turmoil on the board and new challenges for OpenAI’s CEO, Sam Altman, who was fired and rehired in November after a battle with the company’s then-board.
As OpenAI’s business has taken off, Ghodsi, a well-known figure in AI, and his colleagues at Databricks have argued publicly and privately to major software buyers that in the long run, for both performance and security reasons, companies may be better off developing their own AI models or customizing open-source ones rather than getting hooked on OpenAI’s closed-source models. The two companies are on different sides of a brewing debate about the societal and geopolitical ramifications of open-source AI models.
D’Angelo’s approach to Ghodsi may be in keeping with the board’s nonprofit mandate to prioritize limiting societal harms from AI over making money, said a person who has spoken with him. In 2019, when Altman was still a director of the nonprofit, he started the for-profit arm so that OpenAI could raise more money while still adhering to the mission of the nonprofit, which oversees the commercial unit. D’Angelo, whose day job is CEO of question-and-answer site Quora, believes the board should keep that unit—and Altman—in check, said a person who has spoken to him.
But the inherent conflict between the nonprofit’s charter and Altman’s push to build OpenAI’s business while also using his role to pursue deals outside the company—such as AI devices and investing in OpenAI vendors—contributed to Altman’s short-lived ouster. Board directors such as D’Angelo weren’t privy to some of those dealings, multiple people with knowledge of the situation said at the time.
D’Angelo, the only holdover from the OpenAI board that fired and rehired Altman, is now playing a key role in a search for new board members. He, along with two board members appointed following the drama in November, former Treasury Secretary Larry Summers and former Salesforce co-CEO Bret Taylor, have said they are looking to fill at least several open seats. (As part of his deal to return as OpenAI CEO, Altman agreed that he will not regain his seat on the nonprofit board for now.)
Among other candidates, OpenAI board representatives have approached Alexandr Wang, CEO of Scale AI, a firm whose services OpenAI has used to train its models, and former Microsoft and GitHub executive Nat Friedman about potentially joining as directors. The status of those talks also isn’t clear, but neither choice would raise eyebrows among OpenAI and Databricks employees the way Ghodsi’s appointment to the board would. It’s also unclear how advanced the talks between Ghodsi and D’Angelo are.
D’Angelo talked to another OpenAI competitor during the period after Altman was fired and before he was reinstated. D’Angelo talked to Dario Amodei, CEO of OpenAI’s archrival, Anthropic, about possibly becoming OpenAI’s new CEO, The Information previously reported at the time. Those talks didn’t go anywhere, but it showed the lengths D’Angelo would go to secure a future for OpenAI without Altman.
In the meantime, OpenAI is becoming an even more powerful force in the business world. It is now generating more than $130 million in revenue per month. Marquee investors currently value the for-profit entity at $86 billion.
The amount of capital OpenAI may need—Altman has said it might raise another $100 billion on top of the more than $13 billion it has raised already—as well as its fast-growing business and the influence of powerful equity holders such as Microsoft, Sequoia Capital and Andreessen Horowitz could put the for-profit arm at odds with the nonprofit’s mission.
Tensions could also increase if the company’s AI makes more progress following a recent breakthrough. Some OpenAI leaders believe they can develop superintelligent AI that can outperform humans at any task and perhaps help them colonize Mars and beyond. OpenAI says final decisions about whether the company can release new technology, now or in the future, ultimately rest with the nonprofit board, which can “overrule” decisions by leaders of the for-profit arm.
To ease those tensions, OpenAI could rework its structure so the nonprofit entity only owns a small portion of the for-profit arm but receives a percentage of the latter’s profits to fund its mission, a structure similar to those of other nonprofits such as the Lego Foundation, said Josh Wolfe, managing partner at venture firm Lux Capital and a trustee at two nonprofits. (Lux is also a Databricks shareholder.)
For instance, profits from OpenAI could fund AI safety research the board leads or could subsidize access to OpenAI’s technology for students or impoverished communities, Wolfe said. When Altman was reinstated, the new three-person board said it would look at “improving our governance structure” but didn’t imply that it would make any major changes.