Copyright

The authors of original content, such as books, films, and music, are granted an exclusive right to make copies of and perform their content. This right is called copyright. Generative AI (GenAI) often uses content protected by copyright as a training resource when it produces its own content. Some of the content it produces may be original, while some may closely copy the content of a third party. As such, there are many copyright implications for the use of GenAI in an academic setting.

Best Practices When Using GenAI

Create or Supplement Original Content:

Example #1: Use Copilot to create a learning object that tackles your learning goals, for example, a case study in a business class or a short story for an English learning class:

Write a short story using language that a new learner of English could read and comprehend.

The story should incorporate vocabulary related to ordering food from a restaurant, including words such as ‘menu’, ‘price’, ‘tip’, ‘order’, ‘appetizer’, ‘entrée’, ‘server’, etc. The story should take no longer than ten minutes for an average reader to read.

Example #2: Use Copilot to create slide content from your lecture notes:

Generate a slideshow outline to accompany a lecture. The lecture will closely follow the notes that I have provided below: [copy-pasted lecture notes].

Do Not Attempt to Recreate or Adapt Third-Party Materials

Example #1: Do not prompt Copilot with this (an attempt to get access to the NYT’s paywalled article with the headline below):

Finish writing a newspaper article with the headline: How Uvalde’s Newspaper Kept Going, Despite Unimaginable Loss Craig Garnett, the publisher of The Uvalde Leader-News, opens up about covering a tragedy that was — and is — too close to home.

Example #2: Imagine you currently have access to a textbook’s slide decks because it is the required text for your course, however, next term you will lose access to these because your school is switching to a Zero Textbook Cost program. It would be problematic for you to take the texts on the publisher’s slides and prompt Copilot to:

Rewrite the text below so that its content is presented in a new and original way and would not get flagged for copyright infringement: [copy-pasted text].

Note: If you require access to third-party content, contact the Library Learning Commons or the Copyright Services Office for assistance.

Cite All GenAI Outputs

As the legal landscape shifts, it may become necessary to identify what GenAI outputs you have used in your work. Furthermore, we are ethically required to disclose the use of content that is taken from another source, so it is recommended that all uses of GenAI output are accompanied by a citation to identify it as such.

In general, citations should contain information regarding the: developer, date used, name of the tool and its version number, URL, and the prompt used. You may decide for yourself how you would like to cite your use of GenAI, but it is beneficial for students to see you model the guidelines of associations like the APA or MLA.

For use in teaching materials, try to use appendices to document your chat logs, as including prompts in the middle of your materials is disruptive and occupies valuable space. The following example shows how this can be done in MS PowerPoint based on APA: Citing GenAI.

APA Style

Bibliographic Entries

Use the following format:
Developer. (Version Year). Name of GenAI tool (Version, if given) [Type of AI model]. URL
Microsoft. (2024). Copilot [Large language model]. https://m365.cloud.microsoft/chat/

In-Text Citations

In a paper or reading, you should quote your prompt in-text and cite as per standard APA. If your prompt or chat log is long, consider including it in an appendix instead. In a PowerPoint presentation or other document where it would be cumbersome to include the prompt, you can include reference or appendix slides at the end.

Given the prompt “Write a one-sentence story that will shock its readers,” Copilot returned “As the surgeon made the final incision, he realized the patient on the table was his long-lost twin brother” (Microsoft, 2024).

Not all prompts provided to a large language model will produce interesting or thought-provoking writing (Microsoft, 2024; see Appendix A for the full transcript).

Images

Images that you generate with GenAI should include a note under the image in the following format:

Note. Image generated using the prompt “Prompt,” by Developer, Name of tool, Version year (URL).

Figure 1

Robot Composing

Note. Image generated using the prompt "A robot typing at the typewriter, viewed from the side to see the profile of the robot and the typewriter, looking like a photograph," by Microsoft, Copilot, 2024 (https://m365.cloud.microsoft/chat)

MLA Style

Bibliographic Entries

Use the following format:

“Text of prompt” prompt. Name of AI Tool, version number if known, Name of company that made the tool, Day Month Year content was generated, URL.

“Write a one-sentence story that will shock its readers” prompt. Copilot, Microsoft, 1 November 2024, https://m365.cloud.microsoft/chat

In-Text Citations

Cite in standard MLA style, using the prompt as the title of the output. Recall that MLA allows titles to be shortened in-text to improve readability.

“As the surgeon made the final incision, he realized the patient on the table was his long-lost twin brother” (“Write a one-sentence story”).
In “Write a one-sentence story”, we learn that the surgeon’s patient is his twin brother.

Images

Images that you generate should include a caption in the following format:

Fig. No. “Prompt” prompt, Name of tool, Developer, Day Month Year, URL.

Fig. 1. "A robot typing at the typewriter, viewed from the side to see the profile of the robot and the typewriter, looking like a photograph" prompt, Designer, Microsoft, 1 November 2024, https://m365.cloud.microsoft/chat

Copyright Issues in GenAI

Is it legal for GenAI to train on copyright-protected content?
It may be legal. There are various exceptions to copyright in Canada, such as fair dealing. When a person uses copyright-protected content to train GenAI, they may nonetheless be free from copyright liability if the training is a form of fair dealing. Given this uncertainty, it is important to understand the principles of fair dealing and the risks of copyright infringement prior to training your own AI system.
When I use GenAI to create content, do I own it?
It is unlikely that you own it. Copyright is given to human authors to encourage their creative works. Machines such as GenAI do not require these incentives, so they would not receive copyright. However, if a human author uses GenAI outputs in their own work, they do not give up copyright in the parts that they (the human) created. For example, if you author a textbook but use GenAI to create supplemental imagery, you will have copyright in its text but not its images.
Can GenAI outputs infringe copyright?
Yes, GenAI can produce outputs that may infringe copyright. You should never use GenAI to intentionally reproduce copyright-protected content. Furthermore, you must ask yourself if an output resembles something protected by copyright. If you realize that an output would infringe copyright and you proceed to use it, you may be liable for some form of copyright infringement.

Please visit Fanshawe’s Copyright Services Guide for more information on copyright. For questions, contact copyright@fanshawec.ca.