AI corporate governance and Ben & Jerry’s risk
This report analyses the governance structures of OpenAI, Anthropic, and Ben & Jerry’s. It examines the risks of appointing independent guardians to prioritise social missions over shareholder profits. The findings highlight how fully insulated guardians can harm investors and undermine their own missions without proper accountability mechanisms.
Please login or join for free to read more.
OVERVIEW
Introduction
This paper analyses the governance arrangements of OpenAI and Anthropic. These firms raise billions of dollars (p. 1) from investors, then let self-appointed individuals decide how much profit to sacrifice for their mission. Such investor-overriding guardians were previously used at former Unilever subsidiary Ben & Jerry’s. The report suggests that regulators should not count on most firms to voluntarily install such guardians, at least without a kill switch (p. 9).
The Ben & Jerry’s Double-Trouble Meltdown
Ben & Jerry’s was founded in 1978 (p. 10). In 2000 (p. 12), Unilever acquired the company and agreed to create a board controlled by self-perpetuating independent directors to preserve its social mission. Conflicts were managed privately until 2021 (p. 20), when the board announced the non-renewal of the licence held by its Israeli licensee.
This decision generated substantial backlash. During a very short period, Unilever’s share price fell substantially, wiping out almost $10 billion (p. 24) of shareholder value. Over 30 (p. 26) U.S. states had enacted anti-boycott laws, and several divested from Unilever, including the New York State Common Retirement Fund which divested its $111 million (p. 26) of Unilever equity. In 2022 (p. 28), Unilever settled a lawsuit by transferring all rights required to operate the business in Israel to the licensee in perpetuity. The guardians harmed investors and achieved the opposite of their mission, causing double trouble. In 2025 (p. 5), Unilever spun off its ice cream businesses.
OpenAI
In 2019 (p. 36), OpenAI placed its business in a for-profit subsidiary, OpenAI LLC. Microsoft owned 49% (p. 37) of OpenAI LLC and committed up to $13 billion (p. 37) to the entity. However, the LLC was controlled by OpenAI Nonprofit’s directors, who had no obligation other than to carry out the mission.
In 2023 (p. 40), the board attempted to fire CEO Sam Altman over safety concerns. The attempted decapitation nearly wiped out OpenAI. Following pressure, 700 (p. 44) out of OpenAI’s 770 (p. 44) employees signed a letter threatening to leave. Altman was reinstated. Under its updated 2025 (p. 46) structure, OpenAI Group PBC is controlled by OpenAI Foundation, which owns 26% (p. 52) of the PBC. Because the Foundation retains veto rights, the Ben & Jerry’s risk persists.
Should Hershey And Novo Nordisk Reassure OpenAI Investors?
The Hershey Company and Novo Nordisk are both controlled by mission entities run by self-appointed individuals. However, the purpose of their controlling entities is to make money for their beneficiaries, which aligns with investor interests. Because the conflict between guardians and investors is much more severe at OpenAI Group PBC, the experience of Hershey and Novo Nordisk should not provide reassurance.
Anthropic
Anthropic is a PBC controlled by a noncharitable purpose trust. Anthropic most recently raised $30 billion (p. 72). A key difference is that Anthropic’s legal architects installed a kill switch enabling an unspecified super-majority of investors to terminate the guardian arrangement. This makes the guardians only partly insulated, reducing the risk of a meltdown.
Conclusion
The use of fully-insulated guardians creates a risk that they may act in ways that endanger investors and undermine their own mission. Given the fiascos at Ben & Jerry’s and OpenAI, it is unlikely that many other firms will voluntarily experiment with fully-insulated guardians.