Generative Artificial Intelligence Use - Research Policy

Policy Owner: Associate Dean for Research and Discovery

Policy Contact: Kristin Artinger, Associate Dean for Research and Discovery

POLICY STATEMENT

This policy provides guidelines for the effective use of Generative Artificial Intelligence (GenAI) in the dental school research environment so that research integrity, privacy, knowledge dissemination, critical thinking, and innovation can co-exist as GenAI technologies evolve. Considerations for the use of GenAI technologies in data analysis, publication, and grant writing should be made with the following guiding principles in mind.
GenAI tools come with risks including but not limited to accidental disclosure of sensitive data, inclusion of false or endangering content, violation of funder and publisher policies. These risks and strategies to mitigate them are discussed below.

1. Data analysis. GenAI is often used to generate or refine code. This is generally considered to be acceptable if rigorous verification is completed to validate the code. Similarly, image generation and manipulation are prohibited. The use of machine learning is acceptable but requires disclosure. Because of the nature of clinical work, there is a risk of accidental disclosure of sensitive information. Any information entered  in the chat interface, may be used to train the underlying large language model (LLM) that powers the tool and is supplied to answer to another tool user. If this information belongs to another person or organization (e.g., under copyright or license), is subject to regulation (such as PII or HIPAA data), and/or could cause harm to another or the University if publicly released (e.g., proprietary business information), then there is a risk of violation of state and/or federal laws and/or University policies.

   For these reasons, only publicly available information or information you solely own should be used with AI tools, unless otherwise noted by University of Minnesota's University Information Security team and approved for enterprise use. Currently two tools with extra protection licensed by the University: Google Gemini with Enterprise Data Protection and Microsoft Copilot with Enterprise Data Protection. These tools have features to protect your and University data. However, that does not mean they are safe to use with all data. For example, material owned by other organizations and licensed to the University for certain uses (e.g., copyrighted journal articles, book excerpts) should still not be used with these tools.

   Also, the University has different security policies for the two tools, with more data types allowed with Gemini. Student information (FERPA data), personal health information (PHI), and information classified by the University as Private-Highly Restricted information may all be entered into Gemini. However, only the minimum necessary information should be entered and results should only be shared with individuals authorized by the policy governing the data type. PHI has to be explicitly approved and noted in the study by the IRB.

2. Inclusion of false content. Unless you are careful and review and confirm the results supplied by an AI tool it can supply you with false content that is potentially dangerous to your professional and personal reputation and research career. Current AI tools frequently supply content that is copied verbatim from other authors (plagiarism), information that is inaccurate or false, or information that has been made up. If you include any of this in your proposal or publication, you are responsible for it and it could be considered research misconduct. All members of the research team are expected to adhere to this policy. Expectations regarding AI tool use should be discussed openly within research teams to promote transparency and consistent standards.
3. Violation of Funder Policies. Different funders may have different policies regarding the use of AI in preparing grant proposals. The National Institutes of Health (NIH) prohibits “applications that are either substantially developed by AI, or contain sections substantially developed by AI” (see NOT-OD-25-132). Both NIH and the National Science Foundation (NSF) stress that proposers are responsible for all proposal content—whether generated by AI or not—and thus responsible if AI introduces content that violates their policies on research misconduct (which includes the risks above). NSF does not prohibit using AI to prepare proposals, but it encourages (but does not yet mandate) proposers to disclose in the project description if “generative AI technology was used and how it was used to develop their proposal.” Reviewers are prohibited from uploading any confidential grant into a GenAI tool, as it violates confidentiality and integrity rules.
4. Violation of Publisher Policies. Different journals will also have different policies but overall the policy is that AI cannot be an author and authors must take responsibility for the content. If you do use GenAI, it must be disclosed, and must not be used to generate scientific insight, hypothesis or conclusions, because it can fabricate citations, data and results. Generation of figures is not recommended, as you must ensure the accuracy, that supplied data is faithfully represented, and they are free of any copyrighted material and not plagiarized. Also, you should be aware that under current law, figures generated entirely by AI are not eligible for copyright protection.
5. Reviewers are not allowed to upload an unpublished manuscript as it violates confidentiality.
6. Acknowledgement of AI Use. Any of these elements were created in whole or part by AI, or data was analyzed by AI, you should disclose which elements the AI tool created or substantially assisted in analysis, and hope it was verified and is reproduced.
7. GenAI tools such as ResearchRabbit and Elicit are allowed for literature review.

In summary, GenAI can be used as a proofreader or language editor, but the PI must own the scientific content, verify every fact and image, and ensure the proposal is original.

REASON FOR POLICY

The integration of GenAI in research offers significant opportunities for innovation and improvement. When applied to research, technologies such as artificial intelligence have the potential to improve the quality and quantity of data analysis and interpretation.  For dental researchers there is a benefit for efficiency in analysis of high volumes of data.

While the use of GenAI applications (e.g. ChatGPT, Adobe Firefly, Microsoft CoPilot, Google Gemini, ResearchRabbit, Dall-e, etc.) have permeated all facets of research, GenAI presents both opportunities and risks. This policy aims to provide guidelines for the ethical and effective use of GenAI, ensuring that it enhances research while upholding the values of autonomy, ethical decision-making, and critical thinking.

PROCEDURES

There are no procedures associated with this policy.

FORMS/INSTRUCTIONS 

There are no forms/instructions associated with this policy.

APPENDICES

There are no appendices associated with this guideline.

FREQUENTLY ASKED QUESTIONS

There is no FAQ associated with this policy.

DEFINITIONS

External Reviewers

New employees, managers, auditors, IRS, the public

RESPONSIBILITIES

There are no responsibilities associated with this policy. 

RELATED INFORMATION

There is no related information associated with this policy. 

HISTORY

• Approving Body: Council of Chairs
• Date Approved: February 18, 2026
• Date Amended:

ADDITIONAL CONTACTS