What are the risks of AI in philanthropy?

The safeguards the future needs

By Wesley Stevens, Founder of Quillify. October 3, 2026.

Five matter most: floods of low-quality applications, proposals that all sound alike, biased screening, sensitive data leaking into AI tools, and invented facts and citations. Each has a safeguard. Funders can cap volume and do more of the finding themselves. Applicants can draft from their own material and verify every claim. People, with clear accountability, make every funding decision.

Floods of applications

When an application costs little to produce, some applicants will send many. Funders are already responding. The National Institutes of Health now limits each principal investigator to six applications a year. It had seen large numbers of applications from single investigators, some possibly generated with AI tools.

The safeguard: caps where volume is a problem, and funders that do more of the finding themselves, so the best organizations do not have to shout to be heard.

Proposals that all sound alike

Text generated from the same tools tends toward the same phrases. Reviewers who read a round of near-identical prose learn to distrust all of it, including the honest proposals.

The safeguard: drafts built from each organization's own programs, data and voice, with the people who know the work rewriting anything generic. Specific facts are hard to fake and easy to trust.

Biased screening

Automated systems learn from past decisions and inherit their patterns. Detection tools add their own errors: a 2023 Stanford study found that AI detectors flagged most essays by non-native English writers as AI-generated.

The safeguard: people make every funding decision and can explain it, funders publish how their shortlists are built, and no one is rejected on a detector's score alone.

Sensitive data in the wrong place

Grant work touches personal information about clients, donors and volunteers, confidential partner material, and sometimes export-controlled research. Pasting any of it into a public AI tool can expose it.

The safeguard: a written policy on which tools staff may use and what never goes into them, and tools that keep an organization's documents scoped to that organization. How to write that policy.

Invented facts

AI tools can produce figures and citations that look right and do not exist. NIH's own notice warns applicants about fabricated citations.

The safeguard: every number and reference traced to its source before submission. Tools should leave a fact they cannot source for a person to supply rather than invent it.

The quieter risk: rewarding the already visible

A future where funders read records instead of proposals favors organizations with good data. Grassroots groups, new organizations and communities that are hard to measure could fall further behind.

The safeguard: deliberate paths in for those groups, from funders who visit, listen and fund the work that no dataset shows.

Why the safeguards are worth it

Each of these risks has an answer, and the prize is large. Small teams are freed from paperwork, money reaches good work faster, and funders spend their time in communities. Quillify is built around the safeguards that fall to the organization. Drafts come from your own documents and people decide at every step. In reports and letters, missing facts are flagged for you to supply.

What you can do today

More in this series: How will AI change philanthropy? The future of grantmaking, Will AI replace grant writers?, Will there ever be one grant application for every funder?, How will foundations find nonprofits in the future?, What will replace the grant proposal?, What will grantmaking look like without applications?.

Start on the steps that are possible today.

Browsing and keyword search are free, and there is no card required.

Start free