NIH's AI policy, explained from the notices

Release Date:

July 17, 2025

NIH is providing guidance to researchers on the appropriate usage of artificial intelligence (AI) to maintain the fairness and originality of NIH’s research application process. NIH is also instituting a new policy limiting the number of applications that NIH will consider per Principal Investigator per calendar year.

Background

NIH has recently observed instances of Principal Investigators submitting large numbers of applications, some of which may have been generated with AI tools. While AI may be a helpful tool in reducing the burden of preparing applications, the rapid submission of large numbers of research applications from a single Principal Investigator may unfairly strain NIH’s application review processes. The percentage of applications from Principal Investigators submitting an average of more than six applications per year is relatively low; however, there is evidence that the use of AI tools has enabled Principal Investigators to submit more than 40 distinct applications in a single application submission round.

NIH will continue to employ the latest technology in detection of AI-generated content to identify AI generated applications, but it is imperative that all NIH research applications are consistent with the NIH Grants Policy Statement (GPS) Section 2.1.2’s expectation that institutions and affiliated research teams propose original ideas for funding. AI tools may be appropriate to assist in application preparation for limited aspects or in specific circumstances, but researchers should be aware that using AI comes with its own risks. AI use may result in plagiarism, fabricated citations, or other kinds of research misconduct. As a reminder, NIH oversees research misconduct investigations and acts on non-compliance (see GPS Section 4.1.27).

Policy

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. If the detection of AI is identified post award, NIH may refer the matter to the Office of Research Integrity to determine whether there is research misconduct while simultaneously taking enforcement actions including but not limited to disallowing costs, withholding future awards, wholly or in part suspending the grant, and possible termination.

NIH will only accept six new, renewal, resubmission, or revision applications from an individual Principal Investigator/Program Director or Multiple Principal Investigator for all council rounds in a calendar year. This policy applies to all activity codes except T activity codes and R13 Conference Grant Applications. Based on recent data, this limit will affect a relatively small number of Principal Investigators while enabling the NIH to maintain consistently high-quality grant application review and appropriately steward taxpayer dollars.

Effective Date

This policy is effective for applications submitted to the September 25, 2025, receipt date and beyond.

While NIH recognizes that AI tools may be appropriate to assist in limited aspects of application preparation, researchers should be aware that using AI comes with its own risks: AI use may result in plagiarism or possibly other kinds of research misconduct. 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. If NIH identifies evidence of plagiarism, fabrication, or falsification, NIH will refer the matter to the HHS Office of Research Integrity for review.

The purpose of this Notice is to clarify NOT-OD-22-044 on Maintaining Security and Confidentiality in NIH Peer Review: Rules, Responsibilities and Possible Consequences and inform the extramural community that the NIH prohibits NIH scientific peer reviewers from using natural language processors, large language models, or other generative Artificial Intelligence (AI) technologies for analyzing and formulating peer review critiques for grant applications and R&D contract proposals. NIH is revising its Security, Confidentiality, and Non-disclosure Agreements for Peer Reviewers to clarify this prohibition. Reviewers should be aware that uploading or sharing content or original concepts from an NIH grant application, contract proposal, or critique to online generative AI tools violates the NIH peer review confidentiality and integrity requirements.

May 14, 2026

By NIH and HHS Office of Research Integrity Staff

The rapid adoption of artificial intelligence (AI) has led to powerful new advances in biomedical research, from protein folding to cancer diagnosis and more. Recognizing its importance, we have also sought investigator-initiated research involving these tools to address various diseases and conditions affecting the health of Americans (see more in our Highlighted Topics resource). Alongside these NIH-supported scientific advances and priority research areas involving AI, we have also occasionally observed challenges with the use of these tools that affect the integrity of the science we support. Today, we are sharing some helpful reminders on the appropriate use of AI tools when applying, managing awards, and conducting the research process itself.

As a reminder, NIH:

• Limits the number of applications to six that designate the same principal investigator per calendar year, partly in response to the rising numbers of applications generated with AI tools (see these FAQs for more)
• Prohibits peer reviewers from using generative AI for their study section critiques (see these FAQs for more)
• AI tools may be appropriate to assist in application preparation for limited aspects or in specific circumstances, but researchers should be aware that using AI comes with its own risks. Applications that are either substantially developed by AI or contain sections substantially developed by AI are not considered the original ideas of applicants and will not be considered by NIH.

As we explain in this archived NIH Open Mike blog, when we receive a grant application, it is our understanding that it is the original idea proposed by the institution and their affiliated research team. Using AI tools may introduce several concerns related to research misconduct, like including text from someone else’s work without acknowledgment or hallucinating references.

In situations where NIH identifies possible misconduct arising from the use of these tools, NIH and the HHS Office of Research Integrity (ORI) work together per our standard practices. Generally, research misconduct includes fabrication, falsification, and plagiarism. In certain situations, the definition of these behaviors can also be directly related to the inappropriate use of AI. When researchers intentionally, knowingly, or recklessly use AI tools in a manner that deviates from accepted research practices, they may cross the line into research misconduct. And depending on the extent, we may take specific actions to remedy any possible non-compliance.

Situations we may consider:

Situations we may consider:

• Generating data with AI and claiming it was obtained another way, which could constitute fabrication or falsification
• Completely generating grant applications and/or papers using AI tools (including agents) that may have incorrect, misleading, or copied information
• Altering images with AI without full disclosure, which may constitute data falsification
• Presenting AI-generated, non-existent references overtly as real, which could constitute data fabrication (see this 2026 paper)
• Copying substantial portions of text using AI tools without disclosure, which could be plagiarism

What researchers can do:

• Clearly describe in applications, manuscripts, and presentations the use of the AI tools and how they may have been used during the development of an application, research itself, and/or data analysis and subsequent results, including a section on described methods for reproducibility
• Disclose any specific image-editing processes used
• Appropriately cite references and carefully review and confirm information is accurate
• Consult institutional and journal policies and procedures to determine their acceptable practices, including if they allow for disclosing AI use, prohibiting AI-generated images or text, and crediting the tools as authors
• Use the tools to edit or polish text in alignment with institutional, journal, or funders policies

Together, and as part of implementing NIH’s Gold Standard Science Plan, we aim to protect the integrity of the research NIH supports, resulting outcomes, and help foster the responsible use of AI. We do not, however, plan to raise concerns when AI is used in accordance with responsible research practices.

In addition to currently available guidance, ORI is developing best practices to prevent AI-related research misconduct and guidance on the appropriate use of AI tools throughout the research process. If you have any related suggestions, challenges, or opportunities to consider, please reach out directly to [[email protected]](https://grants.nih.gov/cdn-cgi/l/email-protection#bff0edf6fcd0d2d2ccffd7d7cc91d8d0c9). Please include “AI” in the subject line. Your input will help ensure any future guidance is practical and effective for the NIH-supported research community.

Additional Resources:

• Supporting Fairness and Originality in NIH Research Applications
• NIH Strategic Plan for Data Science

National Science Foundation Policy Notice: includes falsification, fabrication, and plagiarism, “whether committed by an individual directly or through the use or assistance of other persons, entities, or tools, including [AI]-based tools.”

Categories: Top Stories

• Topics:
• Grants Policy and Compliance ,
• Research ,
• Research Integrity and Misconduct

The regulatory definition of fabrication pertains to data or results.13 References in publications and grant applications are not generally data or results because they do not arise from a publication’s scientific inquiry.14 Fabricated citations include those generated by AI tools but not reflecting real publications, which are often called hallucinated references. When references do not serve as data or results, fabricated citations may not constitute fabricated data or results.

However, in some cases, such as literature reviews, references are data because they arise from scientific inquiry. In these cases, fabricated citations could be the subject of research misconduct allegations.