Skip to Main Content

Copyright for Faculty and Staff

Librarian

Profile Photo
Sam Cheng
Contact:
FAAD & Copyright Education
Sheridan

Sample Scenarios

Scenario #1 - Analyzing Copyrighted Text

Professor Nevius has designed a new assignment that requires students to leverage Copilot to compare and contrast how personification as a literary device is used in two short stories. Students will first write a brief essay based on selected paragraphs from the stories. After they complete their essay, they copy and paste the same paragraphs into Copilot and prompt the tool to do the same compare and contrast. They then discuss in class how Copilot performs in the analysis and how it compares with their essay. 

Question:

Is it copyright compliant for the students to input the copyrighted excerpts into Copilot? 

Answer: 

This use may fall under fair dealing since the purpose is to review and critique the excerpts and Copilot's performance. Review and criticism are allowed purposes under fair dealing. Only short excerpts, not the stories in their entirety, are used for the intended purpose. Refer to this section for more information about fair dealing. 

Note that some of our licences for library e-resources (e.g., ebooks, journals, and streaming videos) prohibit uploading content into a third-party platform such as an AI tool. Refer to this guide on using library licensed e-resources. 

Generally speaking, it is safer to use Sheridan's licensed Copilot in this situation compared to other AI tools like ChatGPT because Copilot protects inputted prompts and data from being exposed to the public and users have control over retention settings in Copilot. Microsoft also mentions that personal and institutional data is not used to train Copilot's dataset.

Scenario #2 - Remixing Images

Deesha is working on a research assignment exploring AI image generators' possibilities and limits. She plans to upload two images found online into Artbreeder and use the tools available to remix the images and further manipulate the outputs. In a class presentation, she will address the pros and cons of Artbreeder and its implications on an artist's creative process. 

Question: 

Is there a copyright concern for Deesha to use images found online in her research? 

Answer: 

This is a grey area. There are lawsuits related to copyrighted images used to train generative AI tools. For example, a group of artists is suing Stability AI, Midjourney, and DeviantArt for allegedly using their images without permission (read CBR's article). Getty has claimed that Stable Diffusion copied 12 million images in Getty's repertoire to train its AI model without permission (see some of the image outputs in the lawsuit). Interestingly, Getty has recently released their own AI tool claiming that it is free of copyrighted content.  

It is recommended to use copyright free or friendly images to use in generative AI tools. Refer to Sheridan Library's Guide on Finding Creative Commons & Public Domain Images. If an image is Creative Commons licensed, check the licence type to see the conditions of use (e.g., "no derivatives" does not permit remixing or changing content). If you want to use your own images, refer to the terms and conditions of the AI tool you are using to see how your work may be retained and reused. 

Sheridan has a subscription to Adobe Firefly, which you can use to generate images with less copyright concern. Sign in with your Sheridan email address and Single-Sign-On login to access the licensed tool. Firefly is trained on Adobe Stock images as well as openly licensed and public domain content. Adobe states that they are compensating contributors whose content is used in the dataset to train Firefly. Note that Adobe prohibits their stock images and generative AI outputs from being used in other generative AI tools. 

Connect with us