You Need an AI Use Policy Before a Funder Asks for One
Your staff are already using AI. The question is not whether to allow it. The question is whether your organization can describe how it is used, and to whom, when a funder asks.
That question is arriving faster than most policy calendars move. An AI use policy is now part of a complete set of grant policies, sitting alongside procurement and conflict of interest rather than off to the side as a technology matter.
Agencies have started drawing lines
The clearest example so far is NIH. Notice NOT-OD-25-132, "Supporting Fairness and Originality in NIH Research Applications," released July 17, 2025 and effective for applications submitted on or after September 25, 2025, states that NIH "will not consider applications that are either substantially developed by AI, or contain sections substantially developed by AI, to be original ideas of applicants." The same notice limits each Principal Investigator to six new, renewal, resubmission, or revision applications across all council rounds in a calendar year.
The standard is originality, not tool use. NIH did not ban AI tools. It said that work substantially developed by AI is not your original idea. Those are different claims, and the second one is the one your policy has to answer.
NIH is one agency, and the word substantially carries no word count or percentage behind it. There is no government wide rule in 2 CFR 200 on AI in applications, which means expectations are set award by award in the notice of funding opportunity and the terms and conditions. Read for it every time, and write a policy that documents your own practice rather than one that guesses at a threshold nobody has defined.
What the policy has to cover
Six questions, and the answers should fit on two pages.
Your AI use policy
- Which tools are approved, and who approves a new one.
- What never goes into a prompt: client or resident data, personally identifiable information, unpublished budget figures, anything covered by a data use agreement or a confidentiality clause.
- What requires human review before it leaves the building, and who signs off.
- How you disclose AI assistance when an application or a funder asks, and who drafts that language.
- Who is accountable for accuracy in the final document, stated as a person rather than a process.
- What you retain: which prompts and outputs become records, and where they live.
The data question is the one with real exposure. 2 CFR 200.303 requires internal controls that safeguard protected personally identifiable information, and pasting a participant list into a general purpose chatbot to summarize it is a control failure regardless of what your AI policy says. Write that prohibition first and make it specific.
Accountability for what a document says cannot be delegated to the thing that drafted it.
Where the accountability sits
An application is a representation your organization makes to the federal government. That does not change based on what helped produce it. A number that came out of a model and went into a budget narrative unverified is your number once you submit it, and a program design that reads well and does not match what you can actually deliver becomes your performance problem in year one.
So the policy needs a named human at the end of every document. Not a review committee, not a process step. A position that signs.
Two other practical notes. If you buy an AI tool with federal funds, that is a procurement, and it follows your procurement procedure like any other purchase. And if staff use AI on time charged to a federal award, your time and effort records describe the work performed, which does not change because part of the work was drafting with a tool.
Write the policy while the landscape is still forming. An organization that can hand a program officer a dated, specific policy is in a completely different position from one that has to explain what its staff have been doing. Then close the foundation with the registrations, because none of the policy work matters if your SAM.gov registration lapsed in March.
Get practical grant readiness tools and federal funding insights delivered weekly. Subscribe to The Grant Project Newsletter and keep your local government or nonprofit one step ahead of every deadline, rule change, and opportunity.
