Humanity AI's \$18M Pooled Fund Just Reshaped AI Philanthropy. Here Is What Ten Foundations Agreed On — And What They Could Not.

On May 22, 2026, Humanity AI — a collaborative philanthropic initiative organized through Ford Foundation with nine other major foundation co-funders — announced more than $18 million in new grants intended to shape artificial intelligence for the public good. The mechanics: $8 million distributed in $500,000 grants to twelve inaugural grantee organizations, $3 million committed separately to a new AI Civics initiative led by Data & Society with the Digital Public Library of America as anchor partner, and an additional $10 million reserved for a forthcoming open call expected to launch in summer 2026.

The dollar figure is large but not unprecedented. The structural significance is. Humanity AI is the first major pooled philanthropic fund organized around AI's public-interest implications in which ten foundations agreed to write checks into a common pool, agreed on a common set of focus areas, and agreed on a common grantee slate. For nonprofits, advocacy organizations, journalism programs, and academic centers working at the intersection of AI and society, the announcement is also a map. The twelve organizations now in the inaugural cohort represent the strategic priorities that ten of the largest U.S. foundations were willing to align on in 2026. The organizations not in the cohort represent something equally telling.

The Ten-Foundation Coalition

The founding partner list is unusual in its breadth: Doris Duke Foundation, Ford Foundation, Lumina Foundation, Kapor Foundation, John D. and Catherine T. MacArthur Foundation, Mellon Foundation, Mozilla Foundation, Omidyar Network, David and Lucile Packard Foundation, and Siegel Family Endowment. Together these institutions hold combined assets well into the tens of billions of dollars and operate across substantively different programmatic priorities — democracy, the arts and humanities, education, journalism, racial equity, technology policy, and global development.

The list reads as a working alliance between foundations that historically have not funded the same set of issues. Mellon's traditional center of gravity is the humanities and higher education. Doris Duke funds the arts, child well-being, and environmental conservation. Lumina is fundamentally a postsecondary-attainment funder. Packard funds science, conservation, and child development. The unusual proposition Humanity AI is making is that these foundations all see AI as a cross-cutting force that requires shared institutional infrastructure to track, govern, and respond to. The pooled-fund model is the mechanism for sharing that infrastructure.

For mission-driven organizations approaching foundation funders, the implication is significant. The same set of ten funders are now reading the same set of proposals and applying a common set of selection criteria for the AI-and-public-good portion of their giving. For an aligned organization, the pooled model raises the chance that a single high-quality proposal lands in front of multiple funders simultaneously. For an organization whose pitch does not resonate with the pool's framing, the pooled model concentrates risk: a single negative read by the pooled review may produce ten correlated rejections instead of ten independent reads.

The Twelve Inaugural Grantees — And What the Slate Says

The inaugural cohort, each receiving $500,000, is worth reading as a strategic document.

AI Now Institute anchors the labor and political-economy critique of AI, with a sustained research and policy track record on government accountability and the economic consequences of automated systems.

Center for Democracy and Technology sits at the civil-liberties center of the cohort, working on AI governance frameworks that protect civil rights and free expression.

Center on Resilience & Digital Justice brings community-resilience and digital-rights work to communities most affected by automated decision systems.

Council on Foreign Relations LEAD AI represents the foreign-policy and national-security strand of the cohort — informed public discussion and policy analysis on the geopolitical dimensions of AI.

Distributed AI Research (DAIR) Institute anchors the community-driven, grassroots-research alternative to large industrial AI labs, with research methodology that prioritizes the perspectives of affected communities.

Kinfolk Tech sits in the cohort as a creative and cultural-memory project that uses technology, art, and collective practice to surface community narratives.

Partnership on AI is the multi-sector collaborative the cohort uses to bridge into industry; it convenes companies, civil society, and academia around AI deployment practices.

Pulitzer Center anchors the journalism strand, equipping working journalists globally to report on AI systems with technical accuracy.

Student Defense's SHAPE AI initiative focuses on practical guidance for higher-education institutions on responsible institutional AI adoption.

TechEquity brings the workforce and economic-justice perspective, focused on holding the tech industry accountable for the economic harms its products produce.

Data & Society and the Digital Public Library of America jointly lead the separately funded $3 million AI Civics initiative, with Data & Society as the lead and DPLA as anchor partner. The framing is that public libraries and library-based programming are the right institutional site for sustained community engagement with AI.

The shape of the slate is unmistakable. It is a defensive, governance-focused, civil-society-strengthening portfolio. The dominant program categories are democratic-institution safeguarding, worker protection, journalism capacity, public-interest research, community engagement, and AI governance. The cohort is not a portfolio of AI builders. There are no AI startups, no model-development labs, no compute providers, no academic AI research centers in the conventional sense. The selection criteria evidently prioritized organizations whose work is about the regulation, observation, and accountability of AI rather than its production.

That is itself a strategic choice the ten foundations made together. The implicit theory is that the marginal philanthropic dollar in AI in 2026 produces more public benefit when invested in the systems that monitor and shape AI than when invested in AI development itself. Whether that theory is right is one of the open empirical questions of AI philanthropy.

The Absences

The slate is also notable for what is not in it. There are no academic AI safety or alignment research centers in the inaugural twelve. There are no organizations primarily focused on AI-and-climate, AI-and-health, or AI-and-education-delivery. There are no international civil-society organizations from outside the United States, though several grantees work globally. There are no faith-based or religiously affiliated organizations. There are no organizations whose central work is on AI's military applications or autonomous-weapons policy.

Some of those absences will be filled by the $10 million open call expected this summer. Some of them are genuine strategic choices. Either way, organizations whose work falls in those categories should read the announcement as both an opportunity and a constraint: the open call is the path in, and the framing of the inaugural cohort tells applicants what the funders' aperture currently is and where it is not.

The Open Call — Summer 2026

The remaining $10 million is structured as a forthcoming open call. The framing language emphasizes "communities closest to AI's impact" and "bold leaders positioned to advance public interest." That framing suggests a deliberate effort to broaden the cohort beyond the institutionally established organizations in the inaugural slate. The inaugural twelve are, with few exceptions, well-known and well-resourced civil-society institutions. The open call appears designed to reach organizations and leaders that do not yet have established philanthropic relationships with the founding ten funders.

For prospective applicants, the timing is constrained but workable. A summer 2026 launch likely means application materials available in July or August, deadlines in early fall, and grant decisions before year-end. Organizations that are interested should be doing three things now. First, they should monitor announcements from any of the ten participating foundations, since the open call will be promulgated through the existing networks of all ten institutions. Second, they should begin drafting a clear articulation of how their work fits into the inaugural cohort's program framing — democratic institutions, workers' rights, journalism, education, AI governance, public engagement. Third, they should identify which of the ten foundations is the closest fit for their work, since pooled funds often delegate primary stewardship to a single program officer, and warm institutional context will matter when applications are reviewed against a large open-call pool.

What This Means for the Broader Philanthropic Landscape

Humanity AI sits inside a 2026 foundation-funding landscape that has been notably active. The William Penn Foundation's $57.2 million May distribution reflects a major regional funder's response to the same period; UniHealth's $4.57 million approval covered behavioral-health workforce work. What distinguishes Humanity AI from these other announcements is the pooled-fund structure and the explicit cross-foundation alignment on AI as a programmatic priority.

The pooled-fund mechanism is not new in philanthropy — Blue Meridian Partners, Co-Impact, and Lever for Change have all run pooled funds of various designs. What is new is the application of the mechanism to AI policy and AI governance specifically, and at a scale large enough to set a programmatic agenda for a substantial part of the U.S. foundation sector. Watch for two follow-on signals over the next six months. First, whether the open call attracts proposal volumes large enough to suggest that the existing institutional ecosystem is not the binding constraint on AI public-interest work. Second, whether any of the ten participating foundations make separate, additional AI grants that fall outside the pool's framing. If they do, those grants will mark the boundaries of what the pooled framing could and could not accommodate.

For now, the $18 million figure is the headline. The structural fact underneath it — ten major foundations agreeing in writing on what AI philanthropy should be funding in 2026 — is the story that will shape the field over the next three years.