01 · The Question
Can a Funding Agency Decide How Researchers and Reviewers Use AI?
You are preparing a research grant application and want generative AI to improve its clarity, evaluate its methodology, or help draft supporting documents. You have also heard that some funding agencies prohibit AI use during grant review.
Does that prohibition apply to applicants as well? Can a funder reject an application because AI was used to prepare it?
The distinction matters because applicants, peer reviewers, and funding agency personnel have different responsibilities. An applicant ordinarily controls the proposal being developed, subject to institutional and contractual obligations. A reviewer receives confidential material for a specific evaluation purpose and does not have unrestricted authority to disclose it to external systems.
Research funders may therefore establish different AI requirements for proposal preparation, scientific evaluation, administrative processing, and disclosure. Understanding which policy applies to which activity is essential.
03 · What You Need to Know
How Research Funding Organizations Regulate Generative AI
Why Can Research Funders Establish AI-Use Requirements?
Funding organizations establish conditions governing the preparation, submission, evaluation, and administration of research applications. These conditions may be incorporated into funding announcements, application instructions, review agreements, institutional certifications, and award terms.
AI-related requirements may address the authenticity of applicants' intellectual contributions, the reliability of submitted information, the confidentiality of unpublished proposals, or the fairness of evaluation procedures.
Not every policy has the same legal or administrative status. A binding prohibition, an eligibility condition, a disclosure requirement, and a recommendation are different forms of guidance.
Researchers should therefore examine the actual wording and scope of the applicable policy rather than assuming that every statement about responsible AI use constitutes a prohibition.
Applicant AI Use and Reviewer AI Use Are Different Policy Questions
One of the most consequential distinctions concerns who is using the AI system.
| Role |
Typical AI Activity |
Principal Policy Concern |
| Grant applicant |
Drafting, editing, organizing, or evaluating an application |
Originality, accuracy, permitted assistance, and disclosure |
| Grant peer reviewer |
Summarizing, analyzing, or critiquing confidential applications |
Confidentiality, authorized processing, review integrity, and accountability |
| Funding agency personnel |
Administrative screening or research funding operations |
Institutional authorization, information security, fairness, and governance |
A restriction imposed on reviewers does not automatically prohibit applicants from using AI. Conversely, permission for applicants to use certain AI tools does not authorize reviewers to upload confidential applications to the same services.
The relevant policy must be interpreted according to the role and activity it governs.
What Does the NIH Policy Say About Generative AI in Grant Peer Review?
In June 2023, NIH issued Notice NOT-OD-23-149, titled The Use of Generative Artificial Intelligence Technologies is Prohibited for the NIH Peer Review Process.
The notice prohibits NIH scientific peer reviewers from using generative AI technologies to analyze grant applications and formulate peer review critiques. It also explains that uploading or sharing application content, original concepts, or review critiques with online generative AI tools violates NIH peer review confidentiality and integrity requirements.
NIH incorporated this restriction into reviewer confidentiality arrangements. The notice also describes a possible exception for accessibility-related technologies, subject to communication with the designated NIH official before use.
The policy is directed at the NIH peer review process. It should not be interpreted as a universal prohibition on all AI use by researchers preparing grant applications.
Researchers serving as NIH reviewers should consult the current official review agreements and guidance before using any AI-assisted technology in connection with assigned applications.
Does NIH Also Restrict AI Use by Grant Applicants?
Yes. NIH subsequently established a separate policy concerning AI-developed research applications.
In Notice NOT-OD-25-132, issued July 17, 2025, NIH stated that applications, or sections of applications, substantially developed by AI would not be considered the original ideas of the applicants.
The policy applies to applications submitted for receipt dates beginning September 25, 2025. The notice also addresses consequences associated with identifying substantially AI-developed content, including potential post-award enforcement actions.
This is an important distinction from the earlier reviewer-focused prohibition. NIH's policy does not mean that every instance of AI-assisted proofreading or editing is automatically prohibited. Rather, it places particular emphasis on the originality of the applicant's intellectual contribution and the extent of AI development.
Applicants should examine the current NIH notice and any subsequent official clarification rather than attempting to infer a universal permissible percentage of AI-generated text. The policy does not establish a simple numerical threshold for what constitutes substantial development.
AI assistance also does not remove the applicant's responsibility for factual accuracy, originality, or research integrity.
What Does NSF Say About Generative AI in Proposal Preparation and Review?
On December 14, 2023, NSF issued a notice addressing generative AI in its merit review process.
The notice distinguishes between reviewers and proposers.
For reviewers, NSF prohibits uploading proposal content, review information, and related records to non-approved generative AI tools. This restriction reflects the confidential nature of information received during merit review.
For proposers, the notice encourages applicants to indicate in the project description the extent to which generative AI was used and how it contributed to proposal development.
NSF also states that proposers remain responsible for the accuracy and authenticity of their submissions, including content developed with AI assistance.
The language concerning proposer disclosure should be interpreted carefully. An encouragement to disclose is not necessarily equivalent to a universal mandatory disclosure requirement. Applicants must examine the currently applicable Proposal and Award Policies and Procedures Guide, solicitation, and any subsequent policy updates.
How Do NIH and NSF Policies Compare?
| Policy Issue |
NIH |
NSF |
| Generative AI in peer review |
Prohibited for analyzing applications and formulating critiques under NOT-OD-23-149, subject to specified authorization arrangements |
Uploading review materials to non-approved generative AI tools is prohibited |
| AI in applicant preparation |
Substantially AI-developed applications or sections are restricted under NOT-OD-25-132 |
AI assistance is addressed with emphasis on accuracy, authenticity, and disclosure encouragement |
| Applicant responsibility |
Applicants remain responsible for originality and research integrity |
Proposers remain responsible for accuracy and authenticity |
| Confidential review information |
Protected by reviewer confidentiality requirements |
Protected by merit review confidentiality obligations |
This comparison summarizes the cited agency notices and does not replace current agency instructions. Both organizations may issue additional guidance, and individual funding opportunities may impose further conditions.
What Types of AI Restrictions Can a Funding Organization Impose?
Funders may use different regulatory approaches depending on the activity and perceived risk.
-
Prohibition: Certain AI activities may be disallowed, particularly when confidential review material would be processed by unauthorized services.
-
Conditional permission: AI use may be allowed only through approved systems or for specified tasks.
-
Disclosure: Applicants may be required or encouraged to describe how AI contributed to the application.
-
Originality requirements: Funders may restrict applications substantially developed by AI or require applicants to retain responsibility for their intellectual contributions.
-
Integrity requirements: Applicants may remain accountable for fabricated references, inaccurate claims, plagiarism, or other problems regardless of whether AI contributed to them.
These categories can overlap. For example, an activity may be permitted only when appropriate confidentiality protections are in place and any required disclosure has been made.
Does Permission to Use AI Mean an Applicant Can Upload Confidential Material?
No. Permission to use AI for proposal preparation does not automatically authorize uploading every document or dataset to an external service.
A proposal may contain unpublished inventions, confidential partnerships, identifiable participant information, or material governed by contractual restrictions.
Researchers must independently establish whether the proposed processing is permitted under institutional policies, applicable data protection requirements, and the service's actual terms.
The confidentiality implications of uploading an unpublished grant proposal should therefore be considered separately from whether the funding organization permits AI-assisted writing.
What Happens if an Applicant Violates an AI Policy?
The consequences depend on the applicable policy, the nature of the violation, and the funder's administrative procedures.
Possible outcomes may include requests for clarification, findings of noncompliance, refusal to consider an application, or post-award action where the governing policy provides for such consequences.
For example, NIH Notice NOT-OD-25-132 explicitly addresses the treatment of substantially AI-developed applications and potential enforcement consequences when relevant issues are identified after an award.
Researchers should not assume that every minor instance of AI assistance results in disqualification. Equally, they should not assume that an undisclosed or prohibited use is harmless merely because the proposal's scientific content appears sound.
How Should Researchers Verify a Funder's Current AI Policy?
Begin with the official funding organization rather than relying on AI-generated summaries of its policies.
Relevant sources may include the funding announcement, proposal preparation instructions, current policy notices, reviewer agreements, institutional guidance, and formal communications from the program office.
When a policy distinguishes between drafting assistance and substantial AI development, researchers should examine the exact wording and any official examples or clarifications.
When the rule is ambiguous, a written inquiry to the appropriate funding contact may be more defensible than relying on an AI interpretation.
The same principle applies when determining whether AI use must be disclosed in a grant application.
Watch Out
Do not treat a funder's prohibition on AI-assisted peer review as identical to its policy for grant applicants. The two activities involve different permissions, confidentiality obligations, and responsibilities. Verify the rule that applies to your specific role before using an AI system.
06 · What This Means for You
How to Determine Whether Your Intended AI Use Is Permitted
Before using AI in a funding-related activity, identify your role and the specific task. Preparing an application, analyzing a confidential proposal as a reviewer, and performing administrative screening may be governed by different rules.
Next, identify the relevant official policy and determine whether it establishes a prohibition, a condition, a disclosure expectation, or a general recommendation.
A role-based decision framework
If you are preparing your own grant application
Check the funder's applicant policies, originality requirements, disclosure instructions, and institutional rules before using AI.
If you are reviewing a confidential application
Follow the reviewer's confidentiality agreement and explicit AI restrictions. Do not upload proposal material to an unauthorized system.
If you are using AI to draft substantial proposal content
Verify whether the funder restricts substantial AI development and ensure that the applicant's intellectual contribution remains authentic.
If the policy requires or encourages disclosure
Follow the exact applicable instructions and distinguish mandatory declarations from recommendations.
If the policy is unclear or has recently changed
Consult the current official documentation and seek clarification from the responsible funding office before proceeding.
For applicants, it is useful to maintain a record of consequential AI assistance, particularly where disclosure or originality requirements may be relevant. Such documentation should describe what the AI actually contributed rather than merely listing the tool's name.
For reviewers, the priority is protecting confidential application material and following the authorized review procedure. A reviewer should not independently decide that a convenient AI service is sufficiently secure when the funder's policy prohibits its use.
Researchers should also recognize that the applicant's own intellectual contribution is a separate issue from the confidentiality of proposal information.
