The IRB receives many questions about the use of artificial intelligence (AI) tools in research, including for transcription, data analysis, and other data processing activities. Before using any AI tool, study teams should first understand the classification of the data they plan to upload or process. Information about UW–Madison’s current data classification framework is available here. 

Note that the data classification framework will be changing in the near future. The Data Classification Initiative will simplify classifications into three categories and provide clearer examples of the types of data that belong in each category. Data classification helps determine the level of protection required for data and which tools may be used appropriately. Human research data are often classified as Sensitive or Restricted under the current classification framework, and PHI is considered Restricted. Investigators are responsible for classifying their data before determining whether a particular AI tool is appropriate for use. 

UW–Madison’s approved generative AI tools identify the data classifications for which each tool may be used. Currently, Zoom and Webex are the two campus-approved tools that offer AI transcription features. If a study team wishes to use a third-party AI tool for transcription or other research data processing activities not approved by the university, determine whether a cybersecurity risk review is required or consult the Risk Management & Compliance cybersecurity team for help. 

As a reminder, any use of AI in a research study must be described in the IRB application and/or study protocol and receive IRB approval before implementation. 

For more information on IRB considerations in research involving AI, please watch our Clinical Research Forum recording.