A Wall Street Journal piece by Harvard Law School professor Jesse Fried examining the unusual corporate governance structures at OpenAI and Anthropic ahead of anticipated public offerings. Dan Proft welcomed Fried to unpack the argument, framed by the piece’s headline suggesting both companies have effectively placed mission ahead of profit in ways that could create lasting complications for investors.
Fried explained that OpenAI’s structure traces back to its founding in 2015 as a nonprofit by Sam Altman and Elon Musk, motivated in part by a desire to prevent Google from dominating the field of artificial intelligence. As the organization’s capital needs grew, it created a for-profit subsidiary in 2019 to attract venture investment, a subsidiary reportedly on track to go public in 2027. Fried said the resulting structure is unusually complex because the attorneys general of California and Delaware, given the organization’s nonprofit origins, have effectively required that the nonprofit’s board retain authority over appointments to the for-profit entity’s board. Those nonprofit directors, Fried explained, are self-perpetuating, meaning they select their own successors without any direct accountability to investors or a financial stake in the company’s performance, and are explicitly directed to prioritize broadly defined benefits to humanity over profit generation, creating structural potential for conflict with the for-profit business’s commercial objectives.
Fried noted that Musk, who helped found OpenAI’s nonprofit predecessor, has grown critical of how the organization evolved under Altman’s leadership, particularly its shift toward a highly profitable commercial enterprise, and pursued unsuccessful legal action attempting to challenge that transformation. Fried also referenced the brief 2023 ouster of Altman by OpenAI’s board, which was reversed after employees threatened to leave for Microsoft, as an example of how the nonprofit’s governance authority can create instability.
Fried compared OpenAI’s arrangement to a historical precedent involving Ben & Jerry’s, which negotiated a similar structure when it was acquired by Unilever in 2000, embedding independent “mission directors” on its board to preserve the company’s social values alongside its commercial operations. Fried explained that while the arrangement functioned reasonably for years, it eventually led to significant conflict when mission directors pressured the board against renewing the license of Ben & Jerry’s Israeli licensee over its operations in Israeli-controlled territories, triggering divestment actions by state pension funds, including in Illinois, under anti-boycott laws, and ultimately contributing to the resignation of Unilever’s CEO. Fried noted the dispute was eventually resolved by granting the Israeli licensee permanent rights to produce Ben & Jerry’s products, and that Unilever later spun off its ice cream businesses into a separate entity, a move that also had the effect of curbing the mission directors’ influence.
Fried argued that while such governance arrangements may have made sense when OpenAI and Anthropic were operating at a clear technological frontier with few competitors, the rapid emergence of competing Chinese AI firms producing comparable open-source models, along with expanding capabilities at companies like SpaceX, Google, and Meta, has significantly reduced the marginal benefit of restricting only a couple of firms through mission-driven governance. He suggested that in a competitive marketplace with multiple firms pursuing similar capabilities without such restrictions, the practical impact of these accountability structures on broader societal risk may be more limited than originally intended.