Read the Fine Print.

When the company selling the tools is also the one funding their adoption, the question isn’t whether to take the money. It’s what you’re trading for it. I’ve watched enough funding cycles to know that the most dangerous money is the money that arrives when you’re desperate.

The Money

In March 2026, the OpenAI Foundation announced a $1 billion commitment to nonprofit grantmaking.¹ To put that in context: in 2024, OpenAI's total philanthropic grantmaking was $7.5 million. That's not a typo. We're talking about a 130x increase in two years. No foundation in modern history has scaled its grantmaking that fast. Not Gates. Not Ford. Not Rockefeller at its most ambitious. A technology company less than a decade old is now positioned to become one of the largest single funders in the social profit sector.

The People-First AI Fund, the first wave of this commitment, already distributed $40.5 million in unrestricted grants to 208 social profit organizations, with a second wave on the way.² And OpenAI isn't alone. Ten major foundations launched a $500 million Humanity AI initiative to accelerate AI adoption across civil society.³ Add it up: the largest influx of technology-focused philanthropy the sector has ever seen, more money moving faster from fewer sources targeting a single technology platform than at any point in the history of organized giving.

Before I go further, I should be transparent. Two of my clients are among the 208 grantees, and I helped one of them develop their proposal. I’m proud of that work, and I believe in what they’re building. That inside view is part of what qualifies me to write this analysis, and it’s exactly why I want to write it carefully. I’m not arguing against the grants. I’m arguing that the sector’s celebration is happening at the wrong level of analysis.

The press releases write themselves. The LinkedIn posts are flowing. Everyone is celebrating. Surface is where the analysis usually stops. I’ve been in this sector for 35 years, and I’ve learned one thing about large, sudden influxes of money from powerful institutions: the celebration always comes before the reckoning. The question isn’t whether the money is real. It is. The question is what it costs. And the cost isn’t measured in dollars repaid. It’s measured in decisions deferred, in governance never built, in a sector’s technology architecture shaped by the priorities of its largest funder.

The Vacuum

To understand why this money is dangerous, you have to understand the context into which it’s arriving. The social profit sector isn’t just underfunded right now. It’s structurally desperate. And desperate organizations make different decisions than stable ones.

Start with the most visible collapse. USAID has been effectively dismantled. 83 percent of its programs have been canceled. 94 percent of its staff have been laid off.⁴ The New York Times reported in April that Trump's foreign aid overhaul quietly redirected what remained to large U.S.-based contractors, cutting out the local implementing partners who actually do the work on the ground.⁵ A decade of localization rhetoric, the push to shift power and resources to communities closest to the problems, died without a transition plan. Just gone. The infrastructure they built: stranded. The staff they trained: scattered. The communities they served: left to figure it out alone.

Now look at domestic giving. 20 million fewer American households donate today than in 2000.⁶ Donor retention sits at 31.9 percent, meaning for every ten donors an organization acquires, seven will be gone within a year.⁷ One in four Americans plans to cut charitable giving in 2026.⁸ The grassroots giving base is disappearing. Not shrinking. Disappearing.

Meanwhile, 93 percent of high-net-worth donors plan to maintain or increase their giving.⁹ The money isn't vanishing. It's concentrating. And foundations are sitting on record endowments at the five percent legal minimum.¹⁰

SECTOR MATH:  At a five percent payout rate, a $10 billion foundation spends $500 million and keeps $9.5 billion. The endowment isn’t funding the mission. The mission is funding the endowment.

This is the vacuum into which OpenAI’s billion dollars arrives. Not into a healthy ecosystem with multiple funding sources and strong governance infrastructure. Into a sector hollowed out from every direction simultaneously. Government funding gutted. Grassroots giving collapsing. Foundation giving stagnant. And now a single technology company says: We’ll fill the gap. Just use our tools.

The Problem With Vendor Philanthropy

When the company selling the tools is also funding their adoption, the power dynamics shift fundamentally. This isn’t a neutral grant. It’s an investment with a return, the return just isn’t measured in dollars. It’s measured in market share, in platform dependency, in the slow, invisible process of making one company’s technology the default infrastructure for an entire sector.

Apply the Power Map. Who benefits from the People-First AI Fund? OpenAI gets 208 organizations using its tools, building workflows around its platform, generating case studies, and creating dependencies. Those 208 social profit organizations become proof points in OpenAI’s narrative about AI’s transformative potential for social good. Every success story a grantee publishes is a testimonial for OpenAI’s products. Every implementation case study is a sales document wearing a program evaluation’s clothes.

Now apply the Three Questions Test.

Working for whom? For the 92 percent of nonprofits that report using AI, or the seven percent that actually see major impact?¹¹ Says who? The funders commissioning the evaluations, or independent researchers with no financial stake in the outcome? Compared to what? Compared to what those organizations could accomplish if the same money funded core operations, staff salaries, and community-driven program design instead of technology adoption?

The data tells a story the press releases don't. 48 percent of nonprofits report higher costs after adopting AI, not lower.¹² Only nine percent feel prepared to implement AI responsibly. 90 percent of foundations provide no AI implementation support whatsoever.¹³ The sector is being encouraged to adopt a technology that raises costs, that it's not prepared to govern, and for which it receives no implementation support.

SECTOR MATH:  92 percent adoption. 7 percent impact. 48 percent cost increases. That’s not a technology revolution. That’s an efficiency plateau, and into this landscape of underwhelming outcomes, the solution being offered is more money from the company selling the technology.

If a pharmaceutical company funded 208 health clinics to prescribe its drugs exclusively, we’d call it what it is: a conflict of interest. We’d demand disclosure, independent oversight, and patient consent protocols. When a technology company funds 208 social profit organizations to use its tools, we call it innovation. We celebrate it on LinkedIn. The only difference is the industry in which the dependency is being built. We have a word for this in other sectors: vendor lock-in. In the social profit world, we’re calling it a grants program.

I want to be clear: this isn’t about whether OpenAI’s intentions are good. I have no idea what their intentions are, and it doesn’t matter. Intentions are irrelevant when you’re building a dependency architecture. The structure of vendor philanthropy creates capture dynamics regardless of what anyone intends. Good people, working with good intentions, inside a system designed to concentrate power, will concentrate power. I don't mean to sound cynical, but take a step back and look for yourself.

The Harm Reduction Frame

There’s a frame the sector needs that it hasn’t fully adopted. Matthew Maniaci named it directly in a recent thread on Vu Le’s work: “Using technology and engaging in capitalism in general, in modern society, is an exercise in harm reduction.”

That line got to me. It surfaces something I’ve been working with for decades across public health, drug policy, and human rights, without naming it as the through-line of how I think about technology.

Harm reduction has a lineage. It came out of public health responses to HIV/AIDS in the 1980s, when abstinence-only framing of drug use was killing people. The premise is straightforward: people exist inside systems they didn't design and can't fully escape. The question isn't whether to participate. It's how to minimize the harm of participation while working to change the underlying system. You don't refuse to provide clean needles because using drugs is wrong. You provide clean needles because saving lives is right,¹⁴ and you keep pushing for the policy changes that address why people are using in the first place.

The sector’s AI conversation is stuck because it keeps offering two false options. The pro-AI camp wants absolution: adopt the tools, celebrate the efficiency gains, accept the harms as the cost of progress. The anti-AI camp wants purity: refuse the tools, name the harms, treat any use as complicity. Both positions sidestep the actual question: given that none of us is operating outside these systems, how do we minimize the harm we’re complicit in?

Harm reduction asks more of you than either position. It demands you name the harms honestly, including the ones the technology obscures: the data centers in Black and Brown communities, the content moderators with PTSD, the artists losing their livelihoods, the racism encoded in the models themselves.¹⁵ It demands you choose the least harmful option when you have to choose, and be willing to abstain when the math doesn't work. It demands you push for regulation, accountability, and alternatives even while you're operating inside the current system. And it demands you stop pretending that adoption or refusal alone is a complete ethical position.

That’s a much higher bar than what most of the sector is currently clearing. It also happens to be the only ethical posture that takes both harms and operational realities seriously. Vendor-funded AI philanthropy at the levels we're talking about can’t survive that bar. Which is exactly the point.

The Structural vs. Individual Reframe

The sector is framing this moment as an opportunity. Individual organizations can apply. They can compete. They can win funding. The language is constructed at the individual level because individual stories are easier to celebrate. But that’s exactly where the analysis breaks down.

The structural reality operates at a different level. When a single company’s philanthropy can reshape an entire field’s technology choices, governance norms, and implementation standards, we should be concerned. That’s market capture through a grant application. It doesn’t matter that each individual grant looks reasonable. What matters is the aggregate: 208 organizations, across the sector, now building their operations around one company’s technology, funded by that same company’s philanthropy, evaluated by metrics that serve the funder’s narrative. No single grant creates the problem. The system of grants creates the problem.

The sector sees 208 individual grants. I see a system being built, one where the social profit sector’s digital infrastructure is owned, funded, and evaluated by the companies selling the tools. And the organizations accepting the grants aren’t making bad decisions. They’re making the only decision available to them inside a system that has eliminated the alternatives.

The timing isn’t neutral. This money arrives at the exact moment when government funding is being gutted, when grassroots giving is collapsing, when foundations refuse to increase payout rates. The sector isn’t choosing Big Tech philanthropy because it’s the best option. It’s choosing it because it’s the only option left. That’s a monopoly of generosity. And monopolies set their own terms, even generous ones.

What Actually Needs to Happen

I’m not going to pretend the problem is simple. But I will share what I believe the sector needs to do, and what it’s currently failing to do.

Organizations need technology governance before they take the money, not after. If you don’t have a written AI policy covering data ownership, algorithmic transparency, community consent, and exit strategy, you’re not ready to accept vendor-funded technology grants. You’re adding a dependency without a plan for what happens when the funding ends. Governance isn’t a barrier to innovation. It’s the precondition for innovation that doesn’t end in regret. Build the policy before you sign the grant agreement, not after the funder’s tools are embedded in every workflow.

The sector needs collective bargaining with technology funders. Right now, 208 individual grantees are each negotiating alone with a funder whose resources dwarf their own. There's no negotiation; just a terms-of-service agreement everyone clicks through without reading. The sector needs coordinated positions on data rights, platform interoperability, and evaluation standards. One social profit organization can’t negotiate with OpenAI. Two hundred and eight, speaking with one voice, might actually shape the terms. The labor movement worked this out a century ago. The social profit sector hasn’t applied the lesson.

And foundations sitting on record endowments at the five percent legal minimum need to fund the digital infrastructure the sector actually needs, open-source tools, shared technology platforms, independent evaluation capacity, instead of outsourcing that work to the companies selling proprietary solutions.¹⁶ Every dollar a foundation keeps in its endowment while the sector turns to vendor philanthropy subsidizes exactly the dependency I'm describing.

THE QUESTION

If your largest funder’s business model started conflicting with your mission tomorrow, what would you do? And have you built an organization that could actually make that choice?

Sources