PJMF does not accept unsolicited proposals. What it does offer is unusual transparency — a public database of every grant since 2017, updated weekly, which tells you precisely what it funds and how much it writes.
The Patrick J. McGovern Foundation does not take unsolicited proposals. Organizations are invited. That is the first and most important fact, and it means the practical question is not “how do I apply” but “is it worth my time to understand this funder at all.”
For most readers working in AI and social impact, it is — because PJMF does something almost no invitation-only foundation does. It publishes every grant it has made since July 2017 and updates that database weekly. You can see who was funded, for what, and for how much.
That turns an inaccessible funder into a usable research asset. If you want to know what a responsible-AI health project is actually funded at, or which organizations are already working in your niche, the answer is sitting in public.
Secondhand summaries of PJMF — including the earlier version of this page — commonly quote around $35 million in annual grantmaking and $1.6 billion in assets. The filed returns say otherwise.
The FY2023 Form 990-PF (EIN 22-3021586) reports $65.8 million in grants paid against $1.48 billion in total assets and $76.7 million in total expenses. FY2022 was materially the same, at $65.9 million paid on $1.39 billion in assets.
So PJMF gives away roughly twice what it is usually credited with, from a slightly smaller asset base than commonly stated. Both directions matter if you are sizing an ask or assessing whether the foundation is a serious player in your field. It is.
Individual grants run from $30,000 to $2 million, with recent awards clustering between $200,000 and $750,000.
Organizations where AI is decorative. PJMF funds work where data and AI are the mechanism of impact. Bolting a machine-learning component onto an existing program to fit the funder is transparent and does not work.
Anyone hoping to apply this quarter. There is no form. Building the visibility that leads to an invitation is a multi-year activity, not a fundraising tactic.
U.S. domestic service delivery without a technology thesis. The foundation’s taxonomy is specific. Excellent social-service organizations without a data core are outside it.
There is no trick for getting invited, and anyone selling one is selling something. What actually happens is that program staff at foundations like this build their pipelines from research, conferences, co-funders, and referrals from organizations they already fund.
Which means the productive move is to stop thinking about PJMF as a funder to approach and start thinking about it as a community to be visible in. Publish. Present. Collaborate with existing grantees. Show up in the places where the responsible-AI field is having its arguments.
The grants database is where to start, because it is a directory of exactly who those organizations are.
PJMF publishes every grant since July 2017 and updates weekly. Twenty minutes there tells you more about your fit than any amount of reading the mission statement.
PJMF funds organizations where data and AI are the mechanism of impact, not a line item. Adding an AI component to an existing program to attract this funder is legible and does not work.
With no open call, being identified is the pathway. Published work, conference presence, and coalitions with existing grantees are how program staff encounter organizations in this space.
Existing grantees, co-funders and the wider responsible-AI research community are the channels through which invitations actually originate.
PJMF's focus areas are specific and stable. Position against the one you genuinely fit rather than gesturing at several — the foundation reads its own taxonomy carefully.
