Ten Foundations Just Pooled $18M To Decide Who Builds The Public-Interest Side Of AI. The First $8M Went Out By Invitation — But A $10M Open Call Is Coming This Summer.

On May 12, 2026, a coalition of ten foundations operating under the banner Humanity AI announced more than $18 million in pooled grantmaking dedicated to a single proposition: that the public-interest side of artificial intelligence — the part concerned with democracy, labor, journalism, and education rather than with capability and profit — should not be left to the companies building the models. It is one of the clearest signals yet of where institutional philanthropy is placing its bet as AI reshapes civic life, and for mission-aligned nonprofits it is also a roadmap, because the bulk of the money has not yet been awarded.

Of the $18 million, $8 million went to 12 inaugural grantees — each receiving $500,000 — selected by invitation. A further $3 million supports an "AI Civics" initiative led by Data & Society. And the strategically important number: $10 million is reserved for an open call launching summer 2026, with focus areas, timeline, and criteria to be announced in the coming months. That open call is the door most organizations reading this can actually walk through.

Who is behind the money

Humanity AI is a pooled fund — ten foundations putting capital into a common vehicle rather than each running its own program. The collaborative funders are:

• Ford Foundation
• John D. and Catherine T. MacArthur Foundation
• Mellon Foundation
• Mozilla Foundation
• Doris Duke Foundation
• Lumina Foundation
• Kapor Foundation
• Omidyar Network
• David and Lucile Packard Foundation
• Siegel Family Endowment

The mix matters. This is not a single program officer's thesis; it is the convergence of a civil-rights-and-democracy funder (Ford), a media-and-humanities funder (Mellon), an open-internet funder (Mozilla), a higher-education funder (Lumina), an equity-in-tech funder (Kapor), and a science-and-society funder (Siegel) onto one agenda. When ten institutions with different traditions align on the same four priorities, those priorities are worth reading closely — they describe the grantmaking weather for the next several years, not just this round.

Ford's Lori McGlinchey framed the stakes: "This is a generational moment for philanthropy to pool resources and empower talented people working to align technology with our democratic values."

What the first $8 million tells you

The 12 inaugural grantees are the clearest available signal of what Humanity AI considers fundable. They cluster into the collaborative's four declared focus areas — safeguarding democratic institutions, protecting workers' rights, strengthening journalism, and advancing education — and span research institutes, advocacy organizations, and field-building intermediaries:

Two patterns are worth extracting if you are positioning for the open call. First, none of these are AI-capability shops. There is no model-building, no benchmark lab, no compute play. Every grantee sits at the intersection of AI and an existing civic domain — labor, press freedom, libraries, civil rights, foreign policy. The fund is buying domain expertise applied to AI, not AI expertise applied to domains. A workforce nonprofit, a local-news outlet, a public library system, or a civil-legal-aid organization that can articulate a credible AI angle is closer to this fund's center of gravity than a generic "AI for good" startup.

Second, the $500,000 uniform grant size signals a preference for organizations with real operating capacity over experimental micro-grants. These are mid-size, multi-year-capable commitments. An organization with a sub-$250K annual budget and no track record of managing six-figure restricted grants will read as a capacity risk; one that can show prior management of grants at this scale will not.

How to position for the $10 million open call

The open call is the part nonprofits can act on. It has not launched, but the inaugural cohort and the collaborative's stated priorities give enough signal to start now rather than after the criteria drop.

1. Pick your lane against the four priorities. Democracy, workers' rights, journalism, or education. Vague "responsible AI" framing will lose to applicants who name a specific civic harm and a specific population. The funders have told you the four boxes — put your work cleanly inside one.
2. Lead with domain credibility, not AI novelty. The funded cohort proves the thesis: this money rewards organizations that already have standing in a civic field and are extending it to AI. Your strongest sentence is "we have done X for this community for Y years, and here is how AI changes the stakes" — not "we are building an AI tool."
3. Show you can hold a $500K restricted grant. Audited financials, prior six-figure grant management, and a real fiscal infrastructure are table stakes for a pooled fund of this size. If you are early-stage, line up a fiscal sponsor before the call opens.
4. Watch the AI Civics thread. The $3 million to Data & Society and the Digital Public Library of America partnership both point at "community voice in AI development" as a distinct sub-theme. Library systems, civic-tech groups, and community-organizing nonprofits should track whether the open call carves out an AI Civics track.
5. Build the relationship before the deadline. Pooled funds move through known networks. Following the inaugural grantees, showing up in the public-interest-AI field, and being legible to the program staff before applications open materially improves your odds in a competition this visible.

Why a pooled fund behaves differently than a single foundation

It is worth understanding the vehicle, because it changes how you should approach it. A pooled fund is not the same as applying to Ford or MacArthur individually. Ten foundations have agreed on a shared strategy, a shared set of priorities, and — critically — a shared decision process. That has three consequences for applicants. First, the bar for "fundable" is a consensus bar: a project that excites one program officer but worries three others is unlikely to clear, so proposals that are broadly legible across democracy, labor, journalism, and education funders outperform narrow or idiosyncratic ones. Second, pooled funds tend to move in cohorts rather than one-off grants, which is why the inaugural round funded twelve organizations at a uniform amount — expect the open call to behave similarly, selecting a class rather than picking individual winners. Third, the relationships are collective: being known to any one of the ten funders' program staff working on technology and society creates a pathway, because those staff are the people convening around the shared table.

For organizations that have historically chased single-funder relationships, this is a different muscle. The work is less about cultivating one champion and more about being field-legible — visible, credible, and clearly placed within the public-interest-AI ecosystem that the inaugural cohort now defines. That is a months-long positioning effort, not a grant-writing sprint, which is exactly why starting before the summer open call opens is the advantage.

Why this matters beyond the dollars

In federal terms, $18 million is a rounding error. Its significance is directional. As federal research funding tightens and political review reshapes what the government will pay for, philanthropy is stepping into the public-interest-AI vacuum — and ten major foundations have just defined, with real money, what that field looks like. For nonprofits whose missions are about to be reshaped by AI whether they engage with it or not, the Humanity AI cohort is both a warning and an invitation: the funders have decided this work is worth backing, the criteria are becoming legible, and a $10 million door opens this summer. The organizations that study the first cohort now will be the ones ready to walk through it.