The recent decision by Instructure, the parent company of Canvas, to partner with OpenAI and integrate AI into Canvas has sparked confusion and uncertainty among Canvas users, including faculty and instructors who are unsure what this change means for their teaching. In this post, we’ll explain the protocols in place to decide if and when AI features could be made available for opt-in at the University of Chicago. We’ll then lay out what the released AI features in Canvas can do, as well as the controls in place to safeguard data privacy if and when features are made available for you to opt into using.
Protocols for Turning On AI Features in Canvas at UChicago
We would like to stress that Canvas’ AI features will not be abruptly turned on for all users. Instead, certain conditions must be met before we will enable these AI features.
- The Learning Design Team (LDT) in Academic Technology Solutions will review and test any new features. This allows us to raise any concerns if we feel the feature is confusing, problematic to use, or not working as designed. If we feel it is not ready, we will provide feedback to Instructure, and wait for further development work before re-testing.
- IT Security will be consulted to evaluate each AI feature for data privacy and security. The AI features in Canvas must be covered by the appropriate data security and privacy afforded to all content within Canvas, per our license. If it is not explicit, we will ask for additional documentation before allowing use.
- If approved, and the LDT feels the feature is ready for use, we will create documentation, blog posts, and/or workshops to communicate about the new feature(s).
- On the Canvas administrative side, settings will be configured to allow faculty and instructors to opt IN to any AI features that affect courses, rather than being presented with them turned on by default and then opting OUT.
- The activation of AI features in Canvas will be communicated clearly to the University community through various channels, such as Canvas Announcements, knowledge base articles, and blog posts.
AI Controls and Safeguards in Canvas
Privacy Safeguards
Instructure has documented safeguards put in place to protect the privacy of data in your Canvas course when using AI. These safeguards will come into play if and when AI features in Canvas are turned on.
- The AI models used in Canvas are not trained with user data.
- User data is not stored or reused by the models.
- The AI features do not intentionally send personally identifiable information (PII) to the model. However, PII in discussion replies may be unintentionally sent to a model.
Keeping Humans in the Loop
The AI features in Canvas allow you to manage and adjust AI settings. They also keep you in the loop; for example, in AI discussion summaries, the feature allows you to regenerate AI-generated summaries or to fine-tune them by providing further instructions to the AI feature.
For detailed information on the basic functions of Canvas AI and their corresponding safeguards, please see the AI “Nutrition Facts” articles provided by Instructure.
At this time, no AI features have been enabled in UChicago’s Canvas instance.
The currently released AI Features in Canvas (but not yet enabled for us) include:
- Course (Discussion) Translations: AI translates discussion threads into any of ten languages.
- Discussion Summaries: AI Summarization of discussion threads.
- Inbox Translations: AI translates your Inbox messages into any of ten languages.
- Khanmigo Teacher Tools: AI tools created specifically for teachers, through a number of widgets created and delivered by Khan Academy. (Separate account required for users.)
- Smart Search (in Feature Preview): AI is used to interpret search queries to give more relevant results.
More AI features are coming soon to Canvas (at no additional cost), as per the announcement at the annual conference, InstructureCon 2025. Additional advanced AI features will also be forthcoming, specifically IgniteAI’s IgniteAgent, but it will be at an additional cost, as yet unknown. The same protocol will be used to review the upcoming AI features before any decisions are made to make them available in an opt-in way at the course level.
At times, we have asked instructors to help review and test new features in Canvas. If you are interested in giving feedback on AI features in particular, please let us know at academictech@uchicago.edu.
Questions?
If you have questions or concerns about AI features in Canvas, or about any other aspect of teaching with technology, please don’t hesitate to reach out. You can email us at academictech@uchicago.edu, drop by our office hours, or schedule a 1:1 consultation with an instructional designer.
Cover Photo by Tara Winstead