Research Guides: Artificial Intelligence (AI): Copyright and AI

Copyright and Generative AI

Generative AI has raised a plethora of copyright issues and questions. There are different uses of AI and they have different implications for copyright. The United States Copyright Office has issued a three-part report on Generative AI.

The first is on Digital Replicas and whether or not current laws are sufficient (they think not). The second is on Copyrightability which found that current laws are sufficient and that "purely AI-generated material, or material where there is insufficient human control over the expressive elements" is not copyrightable. The third part is on Generative AI Training and discusses whether or not training constitutes copyright infringement. The report states: "To the extent that acts involved in developing and deploying a generative AI model constitute prima facie infringement, the primary defense available is fair use."

Authors Alliance has a very helpful and informative FAQ for authors' and researchers' questions on Generative AI.

Meanwhile lawsuits proliferate. Chat GPT is Eating the World does a good job of keeping up with the latest and updates maps of both international and U.S. lawsuits.

Creative Commons Signals
Creative Commons is working on new preference signals framework designed to increase reciprocity and sustain a creative commons in the age of AI.
Writing About Real People Update: Right of Publicity, Voice Protection, and Artificial Intelligence
Read more about the "right of publicity" issues Generative AI has brought to the fore. These are commonly described as “name, image, and likeness” rights.

This is clearly an evolving space. More clarity on these issues would be helpful. For now, staying informed of the developments and encouraging stakeholders (which is ALL of us!) to voice their perspectives, norms, and concerns is vital to shape how generative AI and copyright will relate and move us towards our common goal of progress in the sciences and useful arts.

As stated above, copyright on generative AI outputs are currently uncopyrightable. But this too is being tested in courts. After all, what determines "insufficient human control over expressive elements?" Where is that line between sufficient and insufficient? Once again we have vagaries. But these things are not clear cut and flexibility in our laws can be beneficial so that there is room for nuance. At any rate, this is—

Not Our First Rodeo

Copyright Law and New Technologies: A Long and Complex Relationship
From the Library of Congress blog: "...technologies continue to raise novel questions about how copyright should apply and whether the law should be revised to fully realize the Founders’ goal that copyright “promote the Progress of Science.”
Copyright Timeline: A History of Copyright in the United States
From the ALA: "Since the Statute of Anne almost 300 years ago, US law has been revised to broaden the scope of copyright, to change the term of copyright protection, and to address new technologies."

Plagiarism, Scholarly Practice Transgressions, or Copyright Infringement?

Copyright is a not always purpose-fit for every issue. Teasing out differences between a copyright infringement, an act of plagiarism, or simply bad scholarly practice can help us understand our use and/or misuse of research materials.

For instance, copying someone else's ideas are not copyright violations (ideas can not be copyrighted) but it could be plagiarism which is considered a violation of an ethical norm. Or, you might attribute an idea properly, but the source is unreliable, in which case you are not practicing good scholarship. If you create something with AI can you verify the sources? Can you properly attribute ideas?

At Smith College we are operating in an ecosystem that must respect and uphold legal, ethical, and good scholarly practice norms. Considering the implications and impact of AI use within our specific educational sphere is essential.

To explore these nuances vis-à-vis AI more deeply see:

Plagiarism, Copyright, and AI
Marc A. Lemley and Lisa Larrimore Ouellette. University of Chicago Law Review Online, forthcoming 2025;
Stanford Public Law Working Paper

What does all this mean for researchers, and people in general, training AI and tweaking AI models? So far, the courts are holding that training generative AI is, as the Copyright Office put forth, a viable fair use. But these cases are still working themselves up through the courts and fair use is always fact-specific (one person's use may be fair while another's is not) so we may not get much guidance on our specific use—and certainly not for several years (courts have their own timescale).

What to do? Assess your fair use! Get comfortable with the purposefully flexible and vague exception that is a cornerstone of the basis of the Copyright Act which calls on us to promote the sciences and useful arts for the sake of progress.

Find out more about fair use:

Copyright and Fair Use
A practical guide to figuring out your use of copyrighted materials.
Putting the ai in Fair Use: Author's Ability to Promote Progress
A Zine about AI and Fair Use by the Author's Alliance.
Copyright and Generative AI: Our Views Today
The always edifying Authors Alliance weighing in on the questions of copyright and generative AI.