Ideas
Philanthropy in the Age of AI
When private capital concentrates around a transformative technology, philanthropy has a specific and irreplaceable role.
Last updated: · 3 min read
By 2027, global private investment in AI will reach $900 billion, with more than 80 percent originating from three countries. That concentration of capital is not just an economic fact. It is a governance fact. And it creates a specific obligation for philanthropy.
Philanthropy has a role here, but only if it's willing to change what it does and how it thinks about its own purpose. The traditional model - fund research, support advocacy, write grants to nonprofits - was designed for a world where the pace of institutional change roughly matched the pace of technological change. That world no longer exists. AI is restructuring public systems - healthcare, education, credit, criminal justice - faster than the institutions responsible for those systems can adapt. Responding to that reality with conventional grantmaking is bringing a filing cabinet to a systems design problem.
I run a $1.5 billion foundation that has committed over $500 million to AI for public purpose. What that work has made clear to me is that the binding constraint on public-interest AI isn't funding. It's institutional capacity - the organizations, the governance infrastructure, the technical expertise, and the human networks that allow democratic societies to exercise authority over AI systems. These don't exist at the scale the moment requires. And markets won't build them, because there's no profit in it. So we build where the market won't go.
Our in-house technical teams work directly with grantees on data governance, model evaluation, and risk assessment, because asking a community health organization to adopt AI without that support is like handing someone keys to a car without teaching them to drive. We created Fund.AI, which brought together more than 150 foundations and unlocked tens of millions in new investment for nonprofits building AI applications. Before that convening, many of those foundations had never made a single AI-related grant. The barrier wasn't interest. It was capacity - theirs and their grantees' - and that gap is the real constraint on getting AI into public-interest hands.
We fund Amnesty International to build tools for monitoring online discrimination and analyzing policy. We fund investigative newsrooms using AI to surface patterns in government records that would take human reporters years to find. We fund Digital Green in India, reaching hundreds of thousands of smallholder farmers with AI-powered agricultural advisory. We fund Khushi Baby, transforming how maternal health data is collected and used in Rajasthan. None of these organizations started as AI companies. They started with deep knowledge of the communities they serve, and we invested in helping them build the technical capacity to use AI in ways that reflect that knowledge. That requires foundations to have technical expertise of their own, to take positions on how systems should be designed, and to stay engaged long after the grant is made.
It also requires engaging with questions of power. The decisions shaping AI aren't being made only in labs and boardrooms - they're being made in legislatures, regulatory agencies, international bodies, and standards-setting organizations. Foundations that stay out of those arenas because they seem "political" are leaving the field to actors with less concern for the public interest. When we supported the UN Secretary-General's High-Level Advisory Body on AI, when we backed the creation of a regional AI center in Trinidad and Tobago, when we partnered with UNESCO to convene 18 Latin American countries on compute infrastructure - those were political acts in the best sense. They were investments in the ability of public institutions to exercise authority over a technology that will shape their citizens' lives for decades.
Philanthropy is also one of the only forms of capital that can think in decades. Venture capital operates on 7-to-10-year fund cycles. Public markets run on quarterly earnings. Building international governance frameworks, training the next generation of people who understand both AI systems and democratic accountability, investing in technical standards that will shape how systems operate long after any particular model is obsolete - none of that pays off quickly, and none of it generates a financial return. That is the work philanthropy exists to do. Whether the sector rises to it will depend on ambition as much as resources.
This progress won't happen through the private market alone. And it won't happen by asking frontline nonprofits to become AI developers overnight. It requires capital that is patient, technically informed, and willing to build things that don't yet exist. The race to build AI has very few participants. The effort to ensure AI serves everyone will require all of us.
Go deeper
Related essays