Imagine a large - no, bigger, much bigger - nonprofit hospital, university, housing developer, or after school program. Bigger by assets than any other. Right now, there are 13 universities in the U.S.A. with more than $10 billion endowments (one of which is a "public" university), with the largest topping $50 billion. Bigger than that.
There is one. OpenAI. Though its size is not based on endowed assets but rather speculative stock value, the organization, which is still as of this writing a nonprofit, is valued at $86 Billion. It's not clear that the organization will continue with its current structure - the events of the last few weeks resulted in a new board and promises to revisit the structure.
Others have written about what the weeks' events mean for the development of AI going forward, the effective altruism (paywall) movement, tech bros, and capitalism. I want to think about what it means - if anything - for civil society.
First, it seems that no one in civil society or the U.S. nonprofit sector really sees the organization as anything other than a commercial firm (it has a capped profit structure, which limits the amount of profit to be returned to shareholders, but only designates profits to be reinvested in the organization (as nonprofits due) after investors are paid out).
I can understand this view, sort of. The sector in the U.S. (as represented by its lobbying/advocacy/infrastructure groups) is still hung up on a certain kind of charitable corporation, designated as 501c3 (OpenAI is such), and doesn't pay much attention to the dozens of other structures that identify as nonprofits. Heck, it's hard to get these groups to address the problematically porous nature of c3s and c4s, they're way behind the eight ball in understanding they swim in a sea filled with informal associations, "Slack"-based "organizations" for mutual aid or volunteering, B corporations, or hybrids. So, perhaps its way too much of an ask to expect recognition among their own of the behemoth of technology development.
Second, the OpenAI events show that the nonprofit governance model is not "strong" enough to outweigh the interests of investors. Given the model's purpose in this situation, and the information that's public, the nonprofit board that fired the CEO was acting as it was intended. I guess no one thought they'd actually do what they were set up to do.
Third, while the argument for data trusts has largely focused on the difference between digital assets and analog ones as the reason for a new organizational form, they're still rare and probably outnumbered by hybrids of profit/non-profit forms. The AI world - especially that which professes some commitment to "ethics", "safety," "responsibility" or "trustworthiness"* - is ripe with hybrids, not trusts. But they're not limited to this field - they're plentiful in journalism, for example. I highlight this in the forthcoming Blueprint 24.
Fourth, it's not just the structure of the organization that matters, it's also the structure of the funding. Many supporters of the AI organizations we captured for our dataset (live link on December 15, 2023) are contributing via deductible donations and commercial investments. The more the donor class uses LLCs and family offices, the harder it is to determine what kind of funding they're putting where. While those who invested for a financial return in OpenAI may be happy with the result of the last few weeks, what about those who donated with an eye on the mission?
Fifth, philanthropy is playing a not insignificant role in these developments. Individuals and organizations associated with effective altruism fund at least 10% of the 160+ AI organizations we track in Blueprint24. They're funding for AI policy fellowships and internships is particularly notable, as these individuals are now well-represented inside policy making bodies. In a very short time, philanthropy has had a significant impact on the development of a major industry, its regulatory overseers (at least in the U.S.A), and the public discourse surrounding it. Had this happened in education, healthcare, or other domains where philanthropy is active we'd see the industry press and professional associations paying close attention (and claiming all kinds of credit). Yet, as noted in the intro, voices in civil society and philanthropy have been awfully quiet about this "impact" on AI.
As someone who has been tracking and explicating the changing nature of organizations in civil society, I see OpenAI as a huge, well-publicized example of something that's been going on for awhile. The nonprofit sector ain't what you think it is. And it's codified boundaries - the legalities that distinguish nonprofit corporations from commercial ones - may not be up to the task of prioritizing mission over financial returns when the assets are digital, the potential for profit so hyped, and the domain (AI development) easy to make seem arcane and "too hard for you to understand" by insiders.
*These are some of the phrases that are being used in the debates over AI development. It's critical to keep an eye on these terms - they don't all mean the same thing, they are used interchangeably though they shouldn't be, and some of them are being used to deliberately gaslight the public about our options when it comes to developing these technologies. Just as political propagandists excel at hijacking terms to denude them of power (see, for example, "fake news"), so, too, do commercial marketers or ideologues excel at using phrases like "safety" to seem universally meaningful, thus providing cover for all kinds of definitions. See Timnit Gebru and Émile Torres on TESCREAL.