September 23, 2026
A donor advisor sits down to draft a quick thank you note. She pulls up the file: a beneficiary’s name, the family’s situation, the specific crisis that made the grant urgent. She pastes it into a personal ChatGPT account to save ten minutes. The note goes out. The task is done.
What she does not know: on that account, the paste can become training data. A company she has never spoken to now holds a stranger’s worst month, permanently, with no one asked and no one answering for it.
That is the fastest way to breach a donor’s trust, a beneficiary’s privacy, and a trustee’s legal duty. It happens in one keystroke, and it happens constantly, because almost nobody in philanthropy has been told where the line sits.
Philanthropy does not run like most industries. It runs on relationships that outlast any single gift, tied to a family’s name, its legacy, its values across generations. A check can be replaced. A reputation cannot.
That makes the work inherently fragile. Every conversation about giving forces people to confront the tangled, often uncomfortable relationship between money and family. Most donors, and the people around them, live with a quiet fear: that they are valued for what they can give, not who they are. That fear puts every relationship on edge.
Trust does not arrive quickly, in philanthropy or anywhere else. It takes years of credibility, humility, and follow-through to build. It takes one bad moment to break, and once broken, it is close to impossible to repair, even when the people responsible meant well.
That is the backdrop against which every AI decision in this sector now gets made.
Used well, artificial intelligence can do real good for this work. It can help a foundation see the disproportionate impact of a policy on a community that a spreadsheet would have missed. It can take the administrative weight off a small nonprofit’s staff so their time goes back to the mission, not the paperwork.
The sector has noticed. The Center for Effective Philanthropy surveyed more than 400 nonprofit and foundation leaders in the spring of 2025 and found that almost two thirds of them now use AI in their work. The same leaders, in the same survey, named their biggest worries without much hesitation: inaccurate results, data security, and bias.
Here is the complication. The gap between what AI can do and what governance has caught up to, that is the problem, and right now almost nobody has been assigned to close it.
We identified seven places where well-meant AI use creates real exposure for donors, foundations, and boards. They show up in how staff handle confidential information, how boards oversee new tools, how contracts get signed, and how federal tax rules for private foundations apply to software nobody thought twice about.
None of these mistakes require bad intentions. Every one of them requires an unwritten policy.
The complete version names all seven mistakes, with the rules behind each one, and closes with a step-by-step fix for individual donors, family foundations, and foundation boards, including a six-step AI governance checklist a board can adopt this quarter.
Download the full article and checklist →
Picture the donor advisor again, the one drafting the thank you note. If her foundation had spent one afternoon setting the account tier that keeps a stranger’s story out of a training set, the note still would have gone out in ten minutes.
Trust took years to build. Protecting it would have taken ten.
Want to know where your organization is exposed? Book a confidential consultation with Leslie Gross →
This post is educational and does not constitute legal, tax, or compliance advice. Confirm every rule cited here with qualified counsel before your board acts on it.