Artificial Intelligence (AI) Frequently Asked Questions

Can WDI employees and other sponsored affiliates use U-M AI?

No. Only U-M faculty, staff, and students have access to U-M AI.

Who sees my prompts and the history of conversations with U-M GPT?

For details on how U-M stores conversations, please review “Who can view my conversation?” on the ITS AI Services FAQ page.

What can you upload to U-M GPT?

Currently, only text may be uploaded to U-M GPT. The limit is approximately 80,000 words uploaded at one time.

How is Zoom AI info collected and shared?

Please review the “Zoom’s statement on data privacy/usage” section of the “Zoom Companion at U-M” knowledgebase article on ITS’ site.

When using DALL-E 3, can I ask it to refine a previously generated image?

No. Each prompt entered is completely new; you cannot ask U-M GPT to refine previously generated images. Please refer to the “AI Model Descriptions” section of the U-M GPT In Depth page for more information on using DALL-E 3.

What does it mean when I prompt U-M AI and receive a “because of content filtering” message?

The service may not respond to a prompt for various reasons (copyright, legal, etc.). For more information on content filter types, please review Microsoft’s Content Filtering page.

What is the Sensitive Data Guide, and what other data types does it include?

The Sensitive Data Guide to IT Services is designed to help you make informed security and compliance decisions about which IT services to use when collecting, processing, storing, or sharing university data. The Examples of Sensitive Data by Classification Level offer additional details on why certain data fits into each category.

When will ITS begin charging for Maizey?

Maizey will remain free through October. Canvas courses continue to receive a $1,000 allocation for the fall semester.

Can MCommunity groups used by Maizey consist of both campus and Michigan Medicine members?

Yes. MCommunity groups can contain a mix of any email domains among its members.

Are U-M AI prompts and responses subject to FOIA requests?

Yes. All U-M AI-generated prompts and responses may be subject to a Freedom of Information Act (FOIA) request. ITS suggests contacting the Office of the Vice President and General Counsel for further details.

I was in a Zoom meeting and a Zoom chatbot entered instead of an attendee. How is that possible?

Whether it announces itself as a chatbot depends on how it was created. You can find more information on Zoom chatbots on the following site: Managing a chatbot.

Can I use U-M AI to complete transcript audits?

Yes, that type of data would fall under moderate sensitivity and you can utilize U-M AI with it.

Will Sora AI become available through U-M AI?

No, there are no plans to add Sora to U-M AI.

How would I cite AI-generated images in marketing or publications?

You should cite all AI-generated images. Please consult the MIDAS guide for additional details.

In Maizey, what does the “Debugging Log Access” setting allow?

This setting allows project owners to provide ITS support staff temporary access to your Maizey to assist with issues. Please consult U-M Maizey: Enable/Disable Debugging Log for more information.

What is the total number of Data Sources I can use in a Maizey project?

You can add 50 data sources to your Maizey project. For more information on data sources and best practices, please check out the following page: U-M Maizey In Depth.

Can I use PowerPoint as a data source in a Maizey?

U-M Maizey will support PPTx files in Google Drive. However, keep in mind that Maizey cannot interpret data in pictures or graphics and will only grab the text out of the file.

Will U-M Maizey support source material in other languages?

We’ve seen sporadic results with other languages. A best practice would be to test as every file and language is different. A colleague just tested a test project he has with converting the source from English to Spanish, and it was still able to answer the questions asked about the source material.

Last Updated on May 10, 2025