What to do if your nonprofit has no AI policy yet
November 19, 2026 · 4 min read
Somebody on your staff pasted a donor list into a chatbot last month, or drafted a grant narrative with one, and nobody can say for sure whether it happened. The board hasn't voted on anything, the executive director has meant to write something for a year, and meanwhile the tools are open in everyone's browser. If your nonprofit has no AI policy, the absence is itself a position, and it's one nobody chose.
When your nonprofit has no AI policy, you still have a position
Valerie Ehrlich and Rachel Kimber, writing for Candid in February 2026, cite a TechSoup benchmark report saying about 80% of nonprofits still lack an AI use policy. Their line for the rest of the post is the useful one: "The absence of a policy isn't the absence of responsibility."
That matters for a board because a policy takes months to write and agree on, and people use these tools this week. A full policy is the right goal. A nine-heading outline can get you there. Until then, you can agree on a small set of rules that cost nothing and can be written on one page.
Five guardrails to agree on now
Candid's post is written for individual staff, and it suggests guardrails each person can set for themselves. A board or director can turn them into shared rules.
- A "never share" list. Name the information that never goes into an AI tool: donor names and gift histories, client or beneficiary details, employee records, and anything covered by a confidentiality agreement. Donor data and AI tools goes through why this decision comes first.
- Check the settings. Candid advises keeping AI "training" turned off and rechecking, because terms can change. Someone should own that check.
- Read the terms of service. The post says to review provider terms regularly to understand what rights you're granting. Pick one person to read the terms of each tool in use, and write down what they find.
- Say what AI isn't for. Candid suggests, as one example, drafting outlines with AI but not writing communications to stakeholders without significant editing. Choose your own line and write it down.
- Keep a log of what works. Document prompts that work, prompts that fail, and uses that could help the organization later. The post says your notes can shape pilot programs and guidelines once a policy is written.
Candid also recommends that staff pay for AI tools rather than use free accounts for work, since paid versions offer stronger data protection. That's a budget decision for your director, and worth a line item if people are already using the tools.
Make it a one-page note, not a policy
Say a staff of eight at a small arts nonprofit has no AI policy. The executive director writes a one-page note with the never-share list, the settings check, and the name of the person who owns each. She sends it to staff, asks for questions, and tells the board at the next meeting that it's temporary. Nothing in it needs a vote, because it restricts risk and doesn't approve new uses.
Keep the note short enough that everyone reads it. Policy nobody knows about is the failure to avoid: a document that exists and that nobody could repeat from memory.
What to do next
Put three dates on a calendar: the day the note goes out, a check about 90 days later on whether anyone has broken it or needed an exception, and the board meeting where a full policy is on the agenda.
If you want to know where your own board and staff stand before you write anything, the AI readiness assessment asks both groups the same questions in private, about seven minutes each, and the report shows where they agree and where they don't. It includes holding off as a legitimate outcome, it's free, and there's no paid version.
The people using these tools today aren't waiting for you, so the first useful thing you can write is short, and the full policy can come after it.