AI Policy Template for Nonprofit Organizations

A nonprofit AI policy template should explain permitted uses, protect donor and beneficiary information, require accurate fundraising claims, and reserve consequential decisions for people. Apply the same boundaries to staff, volunteers, contractors, and board members when they act for the organization.

Use the clauses below as a draft for your nonprofit’s authorized leadership to review and approve. fenbs can retain those approved rules and review responsibilities with shared projects, so the next staff member or connected assistant does not have to reconstruct them from email.

What should you settle before copying the template?

Choose a policy owner and the person or body authorized to approve it under your own governance arrangements. A small organization might assign administration to its executive director and reserve major changes for the board. Do not imply that every nonprofit has to use the same approval structure.

List the work already involving AI, including meeting transcription, donor communications, grant drafting, translation, image creation, and website features. Separate low-risk use of public information from work involving people’s circumstances. A food pantry’s public hours and a family’s intake notes need different treatment.

The broader AI policy guide covers organizational foundations; this page supplies adaptable nonprofit wording. Use your AI risk assessment template to identify uses that need additional review before permission is granted.

What AI-use policy can your nonprofit adapt?

Replace the bracketed fields and check the language against your privacy notices, grant agreements, records rules, employment policies, and service obligations. These are suggested organizational choices. Have the appropriate advisers review any regulated activities rather than assuming charitable status makes the rules identical everywhere.

• Purpose and scope: [Organization] permits approved AI assistance when it supports our mission and respects the people we serve. This policy applies to staff, volunteers, board members, contractors, and others using AI on the organization’s behalf.
• Approved activities: Use only [approved tools and work accounts] for the purposes authorized in [register]. Examples may include drafting from approved public facts, brainstorming event ideas, or organizing nonsensitive action points. Approval does not transfer automatically to a new data type or connected system.
• Protected information: Do not enter donor identities, giving histories, payment details, beneficiary records, safeguarding concerns, sensitive staff information, credentials, or confidential partner material unless the specific use has written approval and appropriate safeguards. Minimize information even in approved workflows.
• Truthful communications: Verify statistics, outcomes, quotations, eligibility claims, financial statements, and descriptions of services against authorized sources. Do not fabricate a beneficiary story, testimonial, photograph, or donation result, or present a synthetic person as a real participant.
• Human review: A named person checks material AI-assisted work before publication, submission, or external use. AI may not independently promise services, accept conditions, send appeals, move money, or change records.
• People and services: Do not use AI alone to determine service eligibility, prioritize beneficiaries, screen volunteers, or make employment judgments. Proposed assistance with such decisions requires separate approval, appropriate review, and a way to raise concerns with a person.
• Transparency and permission: Follow our privacy notices, consent requirements, grant conditions, and publication rules. Explain material AI involvement when appropriate to the communication or required by those obligations. Permission to hold information does not automatically permit sending it to an AI provider.
• Questions, incidents, and exceptions: Ask [contact] before an unclear use. Stop the affected activity and report suspected disclosure or harmful output to [channel]. Exceptions require approval from [authority], with a documented purpose, restrictions, reviewer, and end date.
• Records and review: Keep authoritative work and approval records in approved systems under the retention schedule. [Owner] reviews this policy after a material tool change, incident, or change in obligations and communicates the approved revision to everyone covered.

Attach a short register naming each approved activity, its information boundary, and the reviewer. Give occasional volunteers the same clear instructions as permanent staff. Their personal subscriptions should not become an unreviewed route around the nonprofit’s decisions.

How do you protect donors and the people you serve?

Start with what people were told when you collected their information. The National Council of Nonprofits’ privacy guidance↗ (opens in a new tab) recommends collecting only needed information, honoring privacy policies, explaining AI-related use, and educating staff and volunteers. That is a practical starting point for your own approval questions.

Check the complete input, including attachments, screenshots, recordings, and connected folders. A narrative with names removed can still identify someone through a distinctive situation. Use fictional examples for training, and let the responsible privacy or program lead assess any proposed use of real records.

Retain the approved boundaries in fenbs Decisions and rules, using nonsensitive descriptions. The project needs to know that intake records are excluded; it does not need a sample family’s record to explain why.

What should a reviewer check in fundraising and grant work?

Assign one accountable reviewer for the finished piece, even when several people contribute. Give that person the source for each material number, claim, and quotation. A plausible sentence about meals served or funds raised should never become a substitute for the organization’s actual records.

The IRS explains↗ (opens in a new tab) that exempt organizations must keep records supporting activities, receipts, expenditures, and tax reporting. An AI-written narrative does not replace those records. Keep financial verification with the people and systems responsible for it.

Before a grant submission, check the funder’s instructions, the proposed commitments, the budget, and any disclosure requirement. Before publishing an image or story, check rights and consent. Mark illustrative examples clearly so a generated composite cannot be mistaken for evidence of an actual beneficiary’s experience.

How can you make the rules usable for volunteers?

Give each role a small card of examples: approved activity, forbidden information, required reviewer, and contact for help. Ask a volunteer to explain what they would do with a donor spreadsheet or a request to translate eligibility instructions. Revise any rule that requires them to infer the answer.

Use a fictional event announcement as a first exercise. Check whether the volunteer verifies the location, avoids adding unsupported claims, and obtains approval before publishing. If your nonprofit also commissions software, the separate product requirements guide for AI coding assistants addresses that development handoff.

How to do this with fenbs

1. Create a free account and add a project for your nonprofit’s AI policy review.
2. Add tasks for program, fundraising, privacy, and leadership review. Use fictional scenarios and links to approved policy documents.
3. Record the decisions people approve, including permitted activities, prohibited inputs, and review responsibilities. Keep unanswered exceptions open.
4. Ask a connected assistant to read the project context before policy-related work and save progress with remaining questions and next steps.

fenbs is not a donor database, case-management system, or automatic safeguard against information exposure. Keep sensitive records in the systems your organization has approved. Return to the AI governance hub as you develop related guidance, and start by approving one realistic use well.