OpenAI and Anthropic's mission-director governance structures could cost IPO investors billions, a Harvard Law professor warns.
OpenAI and Anthropic's mission-director governance structures could cost IPO investors billions, a Harvard Law professor warns.

OpenAI and Anthropic are controlled by unaccountable "mission directors" who can prioritize safety and humanity over profits, creating a governance risk unseen in public markets since Ben & Jerry's cost Unilever $10 billion in market value, according to a Wall Street Journal op-ed by Harvard Law School Professor Jesse M. Fried.
"Under the firms' current structures, these directors can't be removed by investors, may have little skin in the game, and can or must ignore profit in decision-making," Fried wrote in the July 16 op-ed. "Investors should scrutinize both companies' arrangements, which may still change before their IPOs, and price shares accordingly."
OpenAI's governance structure is the more dangerous of the two. The company was founded in 2015 as a nonprofit controlled by self-perpetuating directors with a mission to develop AI safely for humanity. In 2019, it created a for-profit subsidiary directly controlled by those same directors. In 2023, the board fired CEO Sam Altman over safety concerns, nearly wiping out investor value before reversing course after most employees threatened to leave for Microsoft. In 2025, regulators refused OpenAI's request to eliminate the nonprofit's control, permitting only a modest restructuring. Today, OpenAI Group PBC — a Delaware public-benefit corporation — remains fully controlled by the nonprofit, whose directors fill the PBC's board and are required to ignore profit entirely on safety and security matters.
Anthropic's structure is less risky but still carries governance concerns. A "purpose trust" controls Anthropic PBC by appointing a majority of its board. Unlike OpenAI's nonprofit, Anthropic's trust can consider investor interests, and its directors can't completely ignore profit on safety issues. The most important safeguard is a kill switch: a supermajority of Anthropic PBC's stockholders can terminate the trust and remove its appointees. But the supermajority threshold hasn't been made public, Fried noted, raising questions about whether the switch is practically usable.
The Ben & Jerry's Precedent
The only comparable case ended in disaster. In 2000, Ben & Jerry's founders sold to Unilever with an arrangement that gave self-perpetuating mission directors control over the brand's social mission. For two decades, conflicts stayed behind closed doors. Then in 2021, the directors blocked a renewal of Ben & Jerry's Israeli licensee deal, triggering a multiyear legal battle. Unilever lost about $10 billion in market value, its CEO resigned, and the company eventually spun off all its ice cream businesses in 2025. The directors, Fried argued, "not only harmed Unilever and its investors but achieved the opposite of their perceived mission."
Why Investors Should Care
The governance risk is particularly acute for OpenAI, which has already experienced its own Ben & Jerry's-style crisis. The 2023 boardroom battle that ousted Altman demonstrated how quickly mission directors can destroy shareholder value. While Anthropic's kill switch provides a theoretical escape valve, the undisclosed supermajority threshold creates uncertainty. Both companies' structures may change before their IPOs, but as Fried noted, the current arrangements demand a risk premium from investors.
Even if the mission directors successfully ensure safe AI development, Fried questioned whether they are necessary. Investor-appointed directors, founders and engineers are unlikely to unleash dangerous technology, he argued, and the Trump administration's recent rapid shutdown of Anthropic's latest models showed the government can act quickly on perceived threats. "Imposing risks on investors might be a small price to pay if OpenAI's and Anthropic's mission directors actually ensure that artificial intelligence benefits humanity," Fried wrote. "But how can they? Even if the directors ensured the two companies create only 'good' AI, others — Google, Meta, SpaceX and their Chinese competitors — can still create 'bad' AI."
This article is for informational purposes only and does not constitute investment advice.