Skip to main content

Using Generative AI While Respecting Privacy

As our community explores artificial intelligence (AI) and begins to adopt generative AI tools (such as ChatGPT or Google Bard), we encourage everyone to try out these tools and see how they can potentially enhance educational, research, and administrative activities at UMD. However, it is critical to recognize that these tools collect, store, and use data you provide–including presenting your content to other users. Providing information to generative AI tools, including queries, writing samples, data sets, and source code, is the same as posting that information on a publicly available website.

We encourage you to explore the potential of generative AI platforms for use in teaching, research, scholarship, and administration as long as no sensitive data are used. When considering what data to put into AI prompts, remember the two Ps. If the information is Public or you have explicit Permission from the person to use their data for that purpose, then it’s OK to use it with AI tools.

Considerations for generative AI use

Some uses, such as asking an AI tool to summarize and grade a student essay submitted for a class assignment violates student privacy. Other examples of instructional data that are not appropriate for use with generative AI include student names and their grades. However, materials you have authored such as course notes or sample essays are fine to use with AI tools. It is also possible, if a student gives explicit permission, to use student-authored materials such as essays or source code with these tools, as long as students have been made aware that they may decline such use without penalty.

In areas of research and administration, some types of data to never put into AI prompts include: information about employees or their performance, research that is controlled or contains Intellectual Property, and papers or grant proposals under review when reviewers are asked to keep materials confidential. However, asking an AI tool to write a review or analyze a published, open access paper or a grant proposal from a public website would be appropriate, since the information is already publicly available. Likewise, asking a tool to write a letter of reference for a fake student with fake qualifications does not raise privacy concerns. But, using the name and information about a real person would not be appropriate.

Data approved for use with generative AI

To determine whether your data requires special attention, consult UMD’s Information Classification Standard (IT-2). If your data is Low Risk (Level 1), it is permissible to upload it to generative AI tools. To process data above Low Risk, any generative AI tool used must have been approved through UMD’s procurement and security review processes. To date, no generative AI tools have been approved for use with data that is greater than Low Risk. If you wish to use a generative AI tool to process data at a level greater than Low Risk, please contact to get help identifying and vetting solutions.

As we all explore the use of generative AI to enhance our work at UMD, we look forward to working with you to employ these tools while continuing to meet our legal and ethical obligations for the data entrusted to us.

-- Adapted from an email to all UMD faculty and staff sent May 26, 2023 and signed by Joseph Gridley, Chief Data Privacy Officer and Jeffrey K. Hollingsworth, VP and CIO --

Back to Top