AI Use Policy and Prompt Checklist for Small Nonprofits

Decide what AI is for

• List the tasks AI may help with, for example first drafts, summaries, translations, spreadsheet formulas and brainstorming.
• List the tasks AI must not do alone, for example final grant budgets, legal or financial advice, and decisions about who receives services.
• Name the tools staff may use, and which account type (personal or organization) they must use.
• Decide whether volunteers follow the same rules as staff.

Keep sensitive data out

• Never paste donor names, contact details or gift histories into a tool your organization has not approved.
• Never paste beneficiary, client, health, child or case information into any public AI tool.
• Remove names and identifying details before asking AI to summarize notes, surveys or stories.
• Check each approved tool's data settings and turn off use of your chats for model training where that option exists.
• Write down where your organization's AI chats are stored and who can see them.

If you would not post it on your website, do not paste it into a tool you have not checked.

Check every output before it leaves the building

• Verify every number, date, statistic and quote against your own records or the original source.
• Open every citation or link the AI gives you; delete any you cannot find.
• Make sure outcomes and impact claims match what your program actually measured.
• Read the draft aloud once: remove phrases your organization would never say.
• Have a second person review anything sent to a funder, the board or the press.

Funders, donors and disclosure

• Check each funder's guidelines for rules on AI-written applications before you use AI on a proposal.
• Never let AI invent beneficiaries, partners, results or budget lines.
• Decide when you will tell readers that AI helped, and write that rule into the policy.
• Keep the final, human-approved version of every proposal and report on file.

Images and real people

• Do not present AI-generated images as photos of real beneficiaries, staff or events.
• Get consent before using a real person's photo, voice or story, whether or not AI edits it.
• Label illustrative AI images clearly when there is any chance of confusion.

Prompts that get usable drafts

• Say who the reader is and what they need to do after reading.
• Paste the facts you want used, and tell the tool to use only those facts.
• Give a length limit and the format, such as a 150-word email or five bullet points.
• Ask the tool to list anything it was unsure about, then check those items first.
• Save prompts that work in a shared document so the team stops starting from zero.

Own and review the policy

• Name one person who owns the AI policy and answers questions about it.
• Agree what to do if sensitive data is pasted by mistake: who to tell and what to record.
• Review the policy and the approved tool list at least once a year, or when a tool changes its terms.
• Share the one-page policy with every new staff member and volunteer who writes on your behalf.