Canvas AI Discussion Summary Tool Input Use

[adrienne_woote1]

Hello!

I can't find any information regarding input/output use with this tool. Are either the prompts entered or the summaries generated used for model training or public use/outputs in any way?

We have been kicking the tires in our college and love this tool, hope to get it out of beta if it ticks all the ethical boxes!

Thanks,

Adrienne

[jkeithle]

Hello!

We have turned this on in Production and would like to share communications with our users about how it works. Does Instructure have any public-facing documentation (FAQ) that we can link to? 

Thank you!

Jules

[WillGHampton]

According to the AI Nutrition Facts for Discussion Summaries, this tool is not trained on user data. The summaries are stored for future use in that same context. In other words, once a summary has been generated, the same summary will be returned unless new responses have been added to the discussion being summarized.  

The Canvas Discussion tool overview and FAQ has more information

[MariSchuldt]

Good question. I recently attended the AI-Enabled University: Innovation, Access and Engagement  breakout session, which was part of the Higher Education Canvas Summit 2025. I believe Zach Pendleton, Instructure's Chief Architect, said that none of the models used for the current AI-driven feature previews can be trained on your institution's data without getting separate signed consent. I wish I could find this writing somewhere so that I could share it with stakeholders at my institution.