A grant from the company selling you the tools is not the same as a grant from a foundation. Five questions to answer before the board votes.

OpenAI’s philanthropic grantmaking went from $7.5 million in 2024 to a $1 billion commitment. Its People-First AI Fund has now moved $49.5 million in unrestricted money, $40.5 million to 208 organizations in December and $9.5 million more in June. Nearly 3,000 organizations applied for the first round. A further $50 million was committed for 2026, and the foundations following OpenAI in are writing their own checks.

Most of that money will do real good. Unrestricted funding at that scale is rare in our sector, and organizations that receive it are right to be glad. With roughly one first-round applicant in 14 funded, the ones who make it are also under real pressure to say yes to whatever comes with it.

The question is not whether to take it. The question is what you are agreeing to alongside it, and whether anyone in your organization has written that down. In our experience the answer is usually no, because the offer arrives as good news and good news does not get due diligence.

Five questions, answered honestly before the board votes, will tell you what you are actually accepting. Answer them before you apply and you will write a better application, because you will know what you are asking for.

1. What happens to this program when the grant ends?

Ask it in dollars, not in spirit. If the grant funds a tool subscription, name the annual cost in year three and the line item it comes from. If it funds a staff position, name who pays for that person when the term is up. If it funds a pilot that works, name what scaling it costs.

A grant that leaves you with a recurring cost and no recurring revenue is a loan with better branding. That may still be worth taking. It is a different decision than the one most boards think they are making, and the difference shows up in year three when the subscription renews and the grant does not.

2. Does the funder benefit if we adopt their product?

Say the answer out loud in the board meeting. If a pharmaceutical company funded 208 clinics to adopt its drugs as the standard of care, we would name that conflict without hesitating. The same structure gets called innovation when the product is software.

Naming it does not mean refusing the money. Plenty of good work gets funded by parties with an interest in the outcome. It means the conflict goes in the minutes, the board knows it exists, and nobody is surprised in two years when a journalist asks about it. Organizations get into trouble over undisclosed conflicts far more often than over disclosed ones.

3. Who owns the data the tool touches?

This is the question most organizations skip and the one with the longest tail. Read the terms for the actual product, not the grant agreement. The grant agreement is written by the philanthropic arm. The terms of service are written by the company.

Four things to establish. Whose servers hold your client records, your donor file, and your case notes. Under what conditions that data can be used to train or improve the model. Whether your staff can turn that off, and whether anyone has. And what happens to the data if you stop using the product.

If you serve people who cannot afford for their information to surface somewhere unexpected, undocumented immigrants, survivors of violence, people in recovery, this question outranks the money. Answer it before the pilot, because after the pilot the data is already in.

4. What are we agreeing not to say?

No grant agreement contains a gag clause. The constraint is softer and more effective than that. Organizations funded by a company become reluctant to criticize that company, and reluctant in a way they will not notice happening. It shows up as a comment letter nobody drafts and a coalition nobody joins.

If your organization does policy or advocacy work that could ever touch AI regulation, data privacy, or labor displacement, decide now, in writing, what you will still say. Then give that document to your board and revisit it when the grant renews. A commitment made before the money arrives is worth considerably more than one made after.

5. Who else is at this table, and are we talking to them?

Two hundred and eight organizations took the first wave and more joined in June. Almost none of them are comparing notes on terms, renewal expectations, data practices, or what actually happened once the money landed and the reporting started.

Individually each grantee has no bargaining power. Collectively they would have a great deal, and the terms that get set in this first period will shape what every subsequent cohort is offered. Find three peer organizations in the same cohort and start a standing conversation. This is the cheapest governance move available to you and the one almost nobody makes.

Where to start

Put the five questions in a one-page memo, answer them in plain language, and attach it to the board resolution accepting the gift. It takes an afternoon.

It is the difference between a board that approved a grant and a board that understood one. If every answer is fine, you have lost an afternoon and gained a document. If one is not, you found it while you still had the option of negotiating.

 

This post is the practical companion to Issue #2 of The Social Prophet, on what the sector is trading for vendor philanthropy. Read it at thesocialprophet.org.

Common Ground Consulting advises nonprofit and foundation leaders on AI strategy, governance, and funding models. If your board is weighing an offer like this, get in touch.