Major public research funders in the United States and Canada keep generative AI out of grant reviewers' hands, but they draw the line in different places. The U.S. National Institutes of Health has barred its peer reviewers from using generative AI since June 23, 2023. The National Science Foundation bars reviewers from uploading proposal material to unapproved tools, and Canada's tri-agency guidance prohibits publicly available online tools for evaluating grant applications.
Key takeaways
- NIH notice NOT-OD-23-149 (June 23, 2023) prohibits generative AI use in NIH peer review, including by advisory council and board members.
- NSF's memo of December 14, 2023, restricts uploads of proposal content to tools not approved by the agency, and allows publicly available information to be shared with current tools.
- Canada's tri-agency guidance calls the use of publicly available online tools for evaluating applications strictly prohibited.
- The stated reason in all three is confidentiality, not the quality of AI writing.
- Applicants face lighter rules: disclosure is encouraged at NSF, and accountability for accuracy stays with the named applicant.
The problem funders are solving
Grant review depends on a promise. Applicants send unpublished ideas, preliminary data, and budgets to a funder, and reviewers agree not to share them. A reviewer who pastes a proposal into a chatbot to summarize it or draft a critique has moved that content to a third party. The funder cannot see where it goes, whether it is stored, or whether it may be reused to improve the tool.
That is the reasoning the agencies give, and it is a different concern from whether AI-written critiques are accurate. Even a tool that wrote excellent summaries would raise the same disclosure problem if it ran outside the funder's control. For that reason the rules read as confidentiality rules that happen to name AI.
United States: NIH
NIH published notice NOT-OD-23-149, titled "The Use of Generative Artificial Intelligence Technologies is Prohibited for the NIH Peer Review Process," on June 23, 2023. According to a June 29, 2023 post by the National Center for Complementary and Integrative Health, the prohibition covers analyzing applications and writing critiques, and it extends to members of National Advisory Councils and Boards. The notice builds on NIH's earlier confidentiality notice, NOT-OD-22-044.
The same post says applicants are not prohibited from using generative AI when preparing an application, but doing so is at their own risk, because tools can reproduce other people's text or invent information. The NIH notice page itself returned an access error to automated requests during the October 2026 check, so the details above rest on the NCCIH post, an NIH institute site.
United States: NSF
The NSF notice to the research community carries a publication date of December 14, 2023, and a last-updated date of May 21, 2026. It says NSF reviewers are prohibited from uploading proposal content, review information, and related records to non-approved generative AI tools, and that doing so would be treated as a violation of the confidentiality pledge reviewers sign. A footnote states that, when the memo was written, the current generation of generative AI did not meet NSF confidentiality standards. Reviewers may share publicly available information with current tools.
Proposers are encouraged, not required, to say whether and how generative AI was used. They remain responsible for the accuracy and authenticity of the submission, and NSF points to its research misconduct rules on fabrication, falsification, and plagiarism.
Canada: the tri-agency guidance
The Canadian Institutes of Health Research, the Natural Sciences and Engineering Research Council, the Social Sciences and Humanities Research Council, and the Canada Foundation for Innovation share one guidance on generative AI in grant proposals. An external expert panel convened in November 2023 published its advice on December 21, 2023, and a public consultation closed on June 14, 2024. The page was last modified on January 6, 2026.
For evaluation, the guidance names ChatGPT and DeepL as examples of tools that may store and reuse submitted text, says uploading application information could breach privacy and intellectual property custody, and calls the use of publicly available online tools for evaluating applications strictly prohibited. For preparation, it recognizes possible benefits, such as helping non-native English and French speakers, while keeping the named applicant accountable for the full contents of the application.
Side-by-side comparison
| Question | NIH | NSF | Canada tri-agency |
|---|---|---|---|
| Reviewers using generative AI | Prohibited for peer review | No uploads to non-approved tools; public information may be shared | Publicly available online tools strictly prohibited |
| Reason given | Confidentiality and integrity of peer review | Confidentiality pledge; information would enter the public domain | Privacy, confidentiality, and loss of intellectual property custody |
| Applicants | Not prohibited, at own risk | Disclosure encouraged; applicant responsible for accuracy | Allowed; named applicant accountable |
| Date on the main document | June 23, 2023 | December 14, 2023 | Page modified January 6, 2026 |
Questions for reviewers and applicants
- Does the funder's current guidance, not an old summary of it, permit any tool for the task at hand?
- Is the content confidential, and would a tool outside the funder's systems receive it?
- Would the tool provider store or reuse the text?
- For applicants: who checks every reference, number, and claim an AI tool helped write?
Rules at these agencies have been revised since 2023 and may change again, so the funder's own notice is the authority for any real submission.
FAQ
Can a grant reviewer use ChatGPT to summarize a proposal?
Not under the NIH, NSF, or Canadian tri-agency rules summarized here. NIH bars generative AI in peer review, NSF bars uploads to non-approved tools, and the Canadian guidance prohibits public online tools for evaluation.
Can an applicant use AI to draft a proposal?
The three funders do not ban it for preparation. NIH describes it as at the applicant's own risk, NSF encourages disclosure, and the Canadian agencies hold the named applicant accountable for the contents.
Why do funders worry about confidentiality more than accuracy?
Proposals contain unpublished work. Once text is uploaded to an outside tool, the funder cannot control how it is stored or reused.
Is there an approved AI tool for reviewers?
NSF's wording refers to tools it has not approved, which implies approved tools could exist. No list of approved tools was found in the sources checked.
Related reading
More in AI Tools and Industry.
Last verified: October 2026
Sources
- Notice to Research Community: Use of Generative Artificial Intelligence Technology in the NSF Merit Review Process, U.S. National Science Foundation, December 14, 2023 (updated May 21, 2026)
- Guidance on the use of Artificial Intelligence in the development and review of research grant proposals, Government of Canada, modified January 6, 2026
- Think again before using generative AI during peer review or as you prepare an application, National Center for Complementary and Integrative Health, NIH, June 29, 2023
- NOT-OD-23-149, The Use of Generative Artificial Intelligence Technologies is Prohibited for the NIH Peer Review Process, NIH, June 23, 2023 (as cited by NCCIH)