Why Humanity AI's First Big Bet Is Libraries: Inside the $2M AI Civics Program and What It Signals for the $10M Open Call This Summer

When the ten-foundation Humanity AI coalition announced its inaugural $18 million grant round on May 12, 2026, the headline most of philanthropy seized on was the sheer collective firepower — Ford, MacArthur, Mellon, Mozilla, Omidyar, Packard, Doris Duke, Lumina, Kapor, and Siegel pooling $500 million over five years to "shape AI for the public good." Twelve organizations received slices of that first $18 million, eleven of them at $500,000 each.

But buried inside that grantee list is the one item that breaks the pattern: a single $2 million award to Data & Society to launch a brand-new program called AI Civics, with the Digital Public Library of America (DPLA) as anchor partner. That award is four times the size of any other inaugural grant. It is not a research grant. It is not a policy advocacy grant. It is, in the language of the funders, a "civic infrastructure" grant — and it is the clearest signal yet about how Humanity AI plans to spend the $10 million it has earmarked for the open call it will launch later this summer.

If you are a nonprofit thinking about applying when that open call drops, AI Civics is the prototype to study.

A Different Theory of the Citizen

Most philanthropic AI-and-society programs over the past three years have been built on what you might call the "consumer protection" model. The user is treated as someone who needs to be educated about AI's risks (algorithmic bias, deepfakes, surveillance), shielded from harms (workplace monitoring, predictive policing), or given tools to use AI more wisely (AI literacy curricula, prompt-craft training, fact-checking apps). The grantee mix in the rest of Humanity AI's inaugural round reflects that lineage: policy shops like AI Now and the Center for Democracy and Technology, civil-rights watchdogs like TechEquity, and journalism-skilling programs at the Pulitzer Center.

AI Civics rejects the consumer frame entirely. In Data & Society's own launch language, the public are not users or victims — they are "civic actors who can and should play a role in steering the course of this consequential technology." The program is being designed to surface and build out "civic tools the public can use to exert political power over AI systems and tech platforms — especially at the community level."

That sounds rhetorical until you look at where the program is being anchored: in the 15,000-plus public libraries that already function as the most-used civic institutions in the country. A 2023 Pew study put the share of Americans who had visited a public library or bookmobile in the prior year at 46 percent. No comparable institution — not town halls, not houses of worship, not community colleges — comes close. DPLA, which already brokers digital collections across more than 7,000 of those libraries, is the connective tissue that lets a national civics program ship through them.

Program director Dr. Meg Young, who built the foundation for AI Civics through the "Understanding AI" event series at the New York Public Library in 2025, frames the design as a deliberate inversion of AI-literacy programs. The goal is not to teach a library patron how to write a better ChatGPT prompt. The goal is to teach a library patron how to find out whether the school district is using an algorithm to assign their kid to a classroom, whether the local sheriff bought a face-recognition system from Clearview, whether the public housing authority is screening tenants through a predictive risk score — and then how to push back through the existing levers of civic life: school-board meetings, public comment periods, FOIA requests, state ombudsman offices, ballot initiatives.

That is a different curriculum than "what is a large language model." It is closer to a 21st-century version of the League of Women Voters voter-guide tradition than to a Coursera AI fundamentals track.

The Two-Year, Two-Phase Architecture

Public materials say AI Civics is funded for two years at the $2 million level. The first phase pairs Data & Society and DPLA exclusively, building out the curricular spine and piloting it through library systems. The second phase expands "sector-by-sector" into labor organizations, faith communities, and educational networks. Both Data & Society and Humanity AI have signaled that subgrants and partnerships outside the founding two organizations are likely once the library phase matures.

That sequencing matters strategically for outside applicants. It is unlikely Humanity AI will fund standalone "AI literacy in libraries" projects through the summer open call, because AI Civics already owns that lane. But it is very likely the open call will fund organizations that can plug into the same civic-actor frame from a different sector: a union local building member-facing tools to audit workplace AI, a state-level coalition translating algorithmic-accountability statutes into community-organizing kits, a Black-church network adapting AI Civics modules for Sunday programming.

In other words, AI Civics functions as the anchor tenant. The $10 million open call will look for the storefronts that complete the block.

What This Says About Humanity AI's Theory of Change

Reading the AI Civics grant alongside the rest of the inaugural round reveals an unstated thesis. The eleven $500,000 grants spread risk across the existing AI-policy-and-research ecosystem — funding organizations that mostly already exist, doing work they were mostly already doing. The single $2 million AI Civics grant funds something new, anchored in non-elite institutions, designed to operate at community scale.

That ratio — roughly $5.5 million to maintain the existing AI-policy field and $2 million to seed a parallel public-facing channel — is the strategic balance the coalition has chosen for year one. It implies that the founders believe the existing field is necessary but insufficient, and that the gap is the absence of any organized constituency that can show up at a city council meeting or a state regulatory hearing and apply political pressure on AI deployment decisions.

The critique inside philanthropy, voiced shortly after the announcement by tech-philanthropy advisor Chantal Forster, is that all eleven of the $500,000 grantees are East-Coast and West-Coast based, with no representation from the heartland or the South. AI Civics is the partial answer to that critique: a delivery vehicle that can reach the 14,000-plus library outlets in rural counties, mid-size metros, and the South, regardless of where the curricular work originates.

If the open call follows the AI Civics template, it will reward applicants who can credibly claim community reach in those geographies — and discount applicants who are essentially research shops or D.C. policy shops looking to extend their existing programs.

What Applicants Should Pre-Position Now

The open call is expected to drop sometime between July and September, with criteria still under development. Five things organizations should be doing now if they intend to compete:

1. Define the civic act, not the educational outcome. Applications that promise to "increase AI literacy among X" will read as duplicative of work already inside the AI Civics lane. Applications that promise to "build the capacity of [community / constituency] to intervene in [specific AI deployment decision]" map to the program's actual theory of change.

2. Bring an anchor institution. AI Civics paired with DPLA because DPLA already has trusted reach into 7,000 libraries. Open-call applicants should partner with an institution that already has presence and trust in the community they intend to mobilize — a union body, a denominational network, a state community-action association, a tribal college consortium. Solo applicants will struggle.

3. Articulate sectoral specificity. The phase-two roadmap names labor, faith, and education as expansion sectors. Applicants from those sectors have an inside track. Applicants from sectors not on the list — health, housing, agriculture, criminal-legal — will need to make a sharper case that their sector is the next obvious expansion point.

4. Show geographic breadth — or geographic depth. The coalition is sensitive to the East-Coast/West-Coast critique. Applicants based outside the coasts, or applicants with credible delivery networks into rural and Southern communities, will be read more favorably. Applicants confined to a single metro need to make a strong argument for that metro as a national model.

5. Treat Rockefeller Philanthropy Advisors as the actual gatekeeper. RPA is the fiscal sponsor and is hiring an Executive Director for the pooled fund. Past Humanity AI program contacts at the founding foundations are less load-bearing than the new RPA team will be. Watch for the Executive Director announcement; that hire's prior portfolio will tell you more about open-call priorities than any of the founding-foundation public statements.

A New Lane for Tech-and-Society Philanthropy

There has been a recurring complaint inside civil-society tech work that funding flows almost entirely to think tanks producing reports nobody outside the field reads, and to advocacy shops fighting battles that resolve before any public is aware they happened. AI Civics is a bet that the bottleneck is downstream of all that: there is no organized public ready to convert a good policy report into political pressure.

Whether the bet pays off will not be obvious for at least two years. But for the next few months, it is the single most informative artifact about what Humanity AI thinks it is buying with $500 million. Anyone planning to apply this summer should read it that way.

Granted's grant detail for the Humanity AI Fund tracks the open-call timing and any sub-grantmaking that emerges from AI Civics partnerships. For ongoing coverage of the broader $500 million coalition, see Granted News.