Generative AI Use Policy


Using generative AI to help with analyzing participants’ responses, or having participants interact with generative AI as part of a study, introduces a number of added risks. These include the sharing of information with a third party not approved to be involved in the research study, increased chance of a participant’s identity being uncovered, the participants’ information being used for non-research purposes (i.e., to further train the AI model), etc. As such, generative AI may only be used to assist with data analysis under situations approved by the IRB. At minimum, this will typically include:

  • Only uploading participants’ anonymous and de-identified responses or information.
  • Not using generative AI when uploading participant’s responses could make them identifiable.
  • Not uploading sensitive information or when public disclosure of the information could pose significant harm (including social, psychological, reputational, financial, legal) to the participant or to FSU.
  • Only using a generative AI with strong security and one with a stated policy that uploaded information will not be used for future training.

To that end, researchers must seek approval before using generative AI, which they would do as part of the regular IRB application (see Section 3. Procedures, and Section 4. Risks and Benefits). In these sections, researchers would describe how exactly generative AI will be used, what information will be uploaded, what steps will be used to ensure all information is anonymous/de-identified, as well as what steps will be taken to minimize the added risk to participants of using generative AI and why the added benefits of using it justify these risks. Additionally, any planned use of generative AI must be described for participants on the Informed Consent document (see Section 7. Consent Forms and Information to Include).

Finally, the Enterprise version of Microsoft 365 Copilot is the default generative AI application that researchers may use. To access it, go to: Microsoft 365 Copilot, and download the app (do not just use the Web version). Once it is installed, open it and log in using your FSU credentials. If done correctly, a green shield with a check mark (Microsoft 365 shield icon) should appear in the top right corner. If it does not, you must stop and contact the FSU Help Desk for assistance.

Researchers who wish to use any generative AI other than Microsoft 365 Copilot - Enterprise must justify this request. Exceptions can be made in certain circumstances, for instance if the researcher has access to a paid version of a different generative AI through a separate institution. However, the researcher must justify why a different application will be used and must provide documentation that its security features and non-training policy are as strong as those of Microsoft 365 Copilot - Enterprise. In general, free generative AI may not be used for participant interactions or for analyzing participants’ non-public data.