Home / AI for Nonprofits / AI policy and data governance for nonprofits

AI policy and data governance for nonprofits

Direct answer. A nonprofit AI policy tells staff which tools they may use, what they may never paste into a public account, and who must review output before it goes into a client record or a funder report. I help teams write one page that legal, programs, and field staff can all defend. It is not a vendor contract and it is not a substitute for counsel on regulated data.

Written by Justin Ingram, healthcare AI consultant. Reviewed November 9, 2026.

Administrative and operational AI only. Not a clinical tool. A person reviews the work.

Why a one page policy beats a binder

Most nonprofits already have acceptable use language for email and laptops. AI adds a new gap because staff can move a case note from an EHR into a personal chat account in seconds. The policy names that move as allowed or forbidden before anyone is embarrassed in front of a board.

I keep the first version short. If nobody reads it, it failed. We add detail only after the first pilot shows a real gap.

Approved tools and accounts

The policy lists organization owned tools, whether a business associate agreement is required for your data class, and whether personal subscriptions are off limits. I do not bless a tool by name on this page. Your counsel and your agreements decide.

Staff need a single place to ask for an exception. That path should name a person, not a generic help desk that never answers.

Client data rules that field staff understand

I write rules as plain do and do not lines. Do not paste full names, dates of birth, member identifiers, or detailed clinical or social histories into tools that are not approved. Do use redacted summaries when a draft would help.

Programs that touch behavioral health or substance use care need tighter lines. Your counsel decides whether Part 2 or other limits apply. I record the rule they give me, not a guess from a blog.

Human review before anything leaves your system

Every policy I write includes a review step. A named staff member checks AI assisted text before it lands in a case file, a letter, or a grant attachment. Review is not optional because the draft sounded confident.

The reviewer signs off with their own judgment. The policy does not shift clinical, eligibility, or legal decisions to the tool.

Disclosure and transparency

Some funders and partners now ask how you use AI in operations. The policy states whether you disclose AI assistance in external reports and how you describe human oversight. I do not promise a funder reaction.

Internal transparency matters too. Staff should know when a workflow uses AI and what the reviewer checks.

Data retention and vendor changes

The policy says what happens to prompts, uploads, and outputs when staff change tools or when a vendor updates terms. I point teams to inventory what is stored where, not to trust a marketing slide.

When a vendor turns on a new feature by default, your policy should say who can opt in and who must wait for review.

How I help you draft and adopt

I interview program, IT, and compliance voices. I merge their lines into one draft. We run one real workflow through the draft before we call it adopted.

Training and the grant narrative should match the policy. If they disagree, staff will follow the shortest path and ignore the rest.

Policy drafting workflow

  1. You send current acceptable use language and a list of tools staff already use, including personal accounts. I read without client content.
  2. We walk one intake note and one funder report path on a call. You correct where reality differs.
  3. I deliver a one page draft with review and data rules. Your counsel marks regulated data language.
  4. A named reviewer tests the draft on one pilot. Staff acknowledge the final version. A human signs off before go live.

Policy checklist for this week

  • List every AI tool staff admitted using in the last month.
  • Write three client fields that never go into an unapproved tool.
  • Name the reviewer for grant and case drafts.
  • Confirm which vendors have signed agreements on file.
  • Schedule counsel to read one page, not a hundred.

Hypothetical: two tools, one confused team

A program director allows ChatGPT on a work laptop while IT blocks the same site on the network. Field staff use phones on personal accounts to draft outreach emails. The policy names one approved path, bans personal accounts for client related text, and assigns review to the communications lead. Nobody is named as a client in this story. It is a composite illustration only.

What this is not

This page is not legal advice, not a HIPAA certification, and not a promise that every vendor will sign your agreement.

Questions people ask

Does a small nonprofit need a written AI policy?

Yes, if staff already paste case or grant text into AI tools. Size does not change the risk. A short policy that people read beats a long policy that sits in a drive folder. I scale the draft to your tool count, not your budget headline.

What should ban personal AI accounts?

If client identifiable information or detailed service histories could leave your systems, personal accounts should be off the table unless counsel approves a narrow exception with a written scope. The policy should say why, not only that IT said no. Staff comply better when they understand the harm model, not only the block screen.

Who should approve the final policy?

Programs, operations, and someone who can speak for privacy agreements. Counsel reviews regulated lines. The executive signs last so the field knows leadership will enforce the same rules. Union or labor relations reviews when bargaining agreements touch tool use. Board members receive a summary, not a last minute surprise.

How does this connect to grants and reports?

Funders care about data handling and whether a human stays accountable. Your policy gives grant writers a consistent sentence about review steps. They can paste that language into data management sections without improvising each notice. It does not guarantee an award.

How often should we update the policy?

Review when you add a vendor, when a tool changes default settings, or when a pilot fails. Annual calendar review alone is too slow for this market. Keep version dates visible on the intranet page staff bookmark and email ops when the version changes.

Related pages

Book a free assessment

Let's Find the Growth Opportunities in Your Practice

Discover how AI can reduce admin, improve patient experiences, and help your practice grow with confidence.

Ready when you are. Free audit

Book audit