OpenAI Is a Cautionary Tale About Nonprofits

Surely you’ve heard by now. On Friday, the board of directors of OpenAI, the world-bestriding startup at the center of the new artificial intelligence boom, fired its chief executive, Sam Altman. He had not been “consistently candid” with the board, the company said, setting in motion a coup — and potential counter-coup — that has transfixed the tech, business, and media industries for the past 72 hours.

OpenAI is — was? — a strange organization. Until last week, it was both the country’s hottest new tech company and an independent nonprofit devoted to ensuring that a hypothetical, hyper-intelligent AI “benefits all of humanity.” The nonprofit board owned and controlled the for-profit startup, but it did not fund it entirely; the startup could and did accept outside investment, such as a $13 billion infusion from Microsoft.

This kind of dual nonprofit/for-profit structure isn’t uncommon in the tech industry. The encrypted messaging app Signal, for instance, is owned by a foundation, as is the company that makes the cheap, programmable microchip Raspberry Pi. The open-source browser Firefox is overseen by the Mozilla Foundation.

But OpenAI’s structure is unusually convoluted, with two nested holding companies and a growing split between who was providing the money (Microsoft) and who ostensibly controlled operations (the nonprofit board). That tension between the nonprofit board and the for-profit company is what ultimately ripped apart OpenAI, because when the people with control (the board) tried to fire Altman, the people with the money (Microsoft) said no. As I write this, Microsoft seems likely to win.

This may all seem remote from what we cover here at Heatmap. Other than the fact that ChatGPT devours electricity, OpenAI doesn’t obviously have anything to do with climate change, electric vehicles, or the energy transition. Sometimes I even have the sense that many climate advocates take a certain delight in high-profile AI setbacks, because they resent competing with it for existential-risk airtime.

Yet OpenAI’s schism is a warning for climate world. Strip back the money, the apocalypticism, the big ideas and Terminator references, and OpenAI is fundamentally a story about nonprofit governance. When a majority of the board decided to knock Altman from his perch, nobody could stop them. They alone decided to torch $80 billion in market value overnight and set their institution on fire. Whether that was the right or wrong choice, it illustrates how nonprofit organizations — especially those that, like OpenAI, are controlled solely by a board of directors — act with an unusual amount of arbitrary authority.

Why does that matter for the climate or environmental movement? Because the climate and energy world is absolutely teeming with nonprofit organizations — and many of them are just as unconstrained, just as willfully wacky, as OpenAI.

Get one great climate story in your inbox every day:

Let’s step back. Nonprofits can generally be governed in two ways. (Apologies to nonprofit lawyers in the audience: I’m about to vastly simplify your specialty.) The first is a chapter- or membership-driven structure, in which a mass membership elects leaders to serve on a board of directors. Many unions, social clubs, and business groups take this form: Every few years, the members elect a new president or board of directors, who lead the organization for the next few years.

The other way is a so-called “board-only” organization. In this structure, the nonprofit’s board of directors leads the organization and does not answer to a membership or chapter. (There is often no membership to answer to.) When a vacancy opens up on the board, its remaining members appoint a replacement, perpetuating itself over time.

OpenAI was just such a board-only organization. Even though Altman was CEO, OpenAI was led officially by its board of directors.

This is a stranger way of running an organization than it may seem. For a small, private foundation, it may work just fine: Such an organization has no staff and probably meets rarely. (Most U.S. nonprofits are just this sort of organization.) But when a board-only nonprofit gets big — when it fulfills a crucial public purpose or employs hundreds or thousands of people — it faces an unusual lack of institutional constraints.

Consider, for instance, what life is like for a decently sized business, a small government agency, and a medium-sized nonprofit. The decently sized business is constantly buffeted by external forcing factors. Its creditors need to be repaid; it is battling for market share and product position. It faces market discipline or at least some kind of profit motive. It has to remain focused, competitive, and at least theoretically efficient.

The government agency, meanwhile, is constrained by public scrutiny and political oversight. Its bureaucrats and public servants are managed by elected officials, who are themselves accountable to the public. When a particularly important agency is not doing its job, voters can demand a change or elect new leadership.

Nonprofits can have some of the same built-in checks and balances — but only when they are controlled by members, and not by a board. If a members association embarrasses itself, for instance, or if it doesn’t carry out its mission, then its membership can vote out the board and elect new directors to replace them. But stakeholders have no such recourse for a board-only nonprofit. Insulated from market pressure and public oversight, board-only nonprofits are free to wander off into wackadoodle land.

The problem is that board-only nonprofits are only becoming more powerful — in fact, many of the nonprofits you know best are probably controlled solely by their board. In 2002, the Harvard political scientist Theda Skocpol observed that American civic life had undergone a rapid transformation: where it had once been full of membership-driven federations, such as the Lions Club or the League of Women Voters, it was now dominated by issues-focused advocacy groups.

From the late 19th to the mid-20th century, she wrote, America “had a uniquely balanced civic life, in which markets expanded but could not subsume civil society, in which governments at multiple levels deliberately and indirectly encouraged federated voluntary associations.” But from the 1960s to the 1990s, that old network fell apart. It was “bypassed and shoved to the side by a gaggle of professionally dominated advocacy groups and nonprofit institutions rarely attached to memberships worthy of the name,” Skocpol wrote.

The sheer number of groups exploded. In 1958, the Encyclopedia of Associations listed approximately 6,500 associations, Skocpol writes. By 1990, that number had more than tripled to 23,000. Today, the American Society of Association Executives — which is, just so we’re clear here, literally an association for associations — counts almost 1.9 million associations, including 1.2 million nonprofits.

This new network includes some nonprofits that claim to have members but are not in fact governed by them, such as the AARP. It includes “public citizen” or legal-advocacy groups, which watchdog legislation or fight for important precedents in the courts, such as Earthjustice, the Center for Biological Diversity, or Public Citizen itself. And it includes independent, mission-driven, and board-controlled nonprofits — such as OpenAI.

There is nothing wrong with these new groups per se. Many of them are inspired by the advocacy and legal organizations that won some of the Civil Rights Movement’s biggest victories. But unlike the member federations and civic associations that they largely replaced, these new groups don’t force Americans to engage with what their neighbors are thinking and feeling. So they “compartmentalize” America, in Skocpol’s words. Instead of articulating the views of a deep, national membership network, these groups essentially speak for a centralized and professionalized leadership corps — invariably located in a major city — who are armed with modern marketing techniques. And instead of fundraising through dues, fees, or tithes, these new groups depend on direct-mail operations, massive ad campaigns, and foundation grants.

This is the organizational superstructure on which much of the modern climate movement rests. When you read a climate news story, someone quoted in it will probably work for such a nonprofit. Many climate and energy policy experts spend at least part of their careers at some kind of nonprofit. Most climate or environmental news outlets — although not this one — are funded in whole or part through donations and foundation grants. And most climate initiatives that earn mainstream attention receive grants from a handful of foundations.

There is nothing necessarily wrong with this setup — and, of course, an equivalent network devoted to stopping and delaying climate policy exists to rival it on the right. But the entire design places an enormous amount of faith in the leaders of these nonprofits and foundations, and in the social strata that they occupy. If a nonprofit messes up, then only public attention or press coverage can right the ship. And there is simply not enough of either resource to keep these things on track.

That leads to odd resource allocation decisions, business units that seem to have no purpose (alongside teams that seem perpetually overworked), and decisions that frame otherwise decent policies in politically unpalatable ways. It regularly burns out people involved in climate organizations. And it means that much of the climate movement’s strategy is controlled by foundation officials and nonprofit directors. Like any other group of executives, these people are capable of deluding themselves about what is happening in the world; unlike other types of leaders, however, they face neither an angry electorate nor a ruthless market that will force them to update their worldview. The risk exists, then, that they could blunder into disaster — and take the climate movement with them.