Skip to Main Content

Artificial Intelligence

Disclaimer

The information presented by the Library about copyright is intended for information purposes, and should not be construed as legal advice. While we cannot provide legal advice, we can help explain these issues in greater detail.

Copyright Law and AI

In very simplified terms, legal questions around AI and copyright fall into one of two categories.

Inputs

Is it legal to train AI/language models using copyrighted works without the copyright holder’s explicit permission?

  • Like many things in life, it depends. To date, several high-profile cases have deemed that training AI using copyrighted material is permissible under fair use doctrine (Bartz v. Anthropic, Kadry v. Meta). These rulings follow previous legal precedent about Text and Data Mining – a research method that interacts with copyrighted works in a similar way to AI model training. Specifically, courts have ruled that AI engines make “transformative” (i.e. new and different) use of copyrighted works, as opposed to superseding the original works by serving the same purpose and/or creating a market replacement for any of the copyrighted works upon which they are trained. Transformative use weighs heavily in favor of fair use in copyright disputes.
  • However, a caveat that potential AI trainers must consider is the means by which they gain access to copyrighted materials. For example, although Anthropic’s use of copyrighted works to train its Claude AI tool was deemed a fair use in court, its acquisition of many of those copyrighted works through pirating sites like LibGen to create a training library was likely to be deemed illegal, barring Anthropic's agreement to settle out of court. To clarify: if someone illegally accesses a copyrighted work, that does not prevent the user from potentially making fair use of that work; but, regardless of whether the use is fair or not, the user is not protected from the consequences of illegally accessing the work in the first place. The access of the work and the use of the work are considered and treated differently in court.

Do generative AI companies gain copyright over your prompts?

“OpenAI currently claims no rights in the inputs you might submit into their system or outputs generated by OpenAI services. The terms also inform users that they may choose to opt out of OpenAI using their Content for training AI. However, even if you opt out of AI training, their terms also mean that OpenAI can make any internal use of your Content (including inputs and outputs). It is highly likely that anything you upload or otherwise submit to OpenAI may be used in ways that might not be easy to identify at this time, even if it is not directly training AI.”

  • Licensing or contractual terms that prohibit training AI programs with published materials would likely prohibit including those materials as part of prompts. See more about how this limits training using UNCW Library materials in the next section below.

Outputs

Can you copyright a work that has been created with the assistance of AI?

  • Surprise! It depends. Specifically, much of a work’s copyrightability depends on the degree of human involvement in its creation. Even before AI was center stage, U.S. courts decided in Naruto v. David Slater that non-human actors could not be copyright owners. As such, the results of a single prompt to ChatGPT would likely not be copyrightable. But, the more a human edits, revises, arranges, or otherwise creatively alters an AI output, the more likely it is to be copyrightable. The U.S. Copyright Office has indicated the copyrightability of such works would be handled on a case-by-case basis.
  • Were the creator of an AI-assisted work subject to a copyright infringement suit, the primary question would be whether the new work was a substantially similar replacement to the original work, or if it was transformative in some way so as to be covered by fair use. That said, because AI users generally do not have direct access to the materials AI was trained upon, it would be unlikely for a copyright suit to be brought against a user, and more likely for a suit to be brought against an AI program’s parent company (Authors Alliance, 2025).

 

For the most up-to-date guidance on issues around AI and copyright, we recommend Authors Alliance’s Generative AI FAQ and the U.S. Copyright Office’s page on Copyright and Artificial Intelligence.

 

UNCW Library Materials and AI

Copyright law gives you certain rights, including fair use for research and education.

Contract law can override those rights depending on the specific terms of the contract.

Can I use UNCW Library materials in generative AI prompts, or to train an AI program/language model?

  • Publishers are copyright owners. They often hold copyright over very large collections. Libraries make contractual agreements with publishers for electronic subscription access to works for which publishers hold copyright. The Library subscribes to, but does not own, this content. When UNCW Library signs a licensing agreement with a publisher so you can access and use the publisher’s content, both the University and its users (that’s you) must comply with those contract terms.
  • If you’re using a corpus of materials from a Library-licensed database or other Library-licensed online resource, there may be contractual arrangements or terms of service in the license agreements that the Library has signed, or in the source website’s terms of use, that override your ability to rely on fair use of that licensed content, such as using it to train language models.

  • The following kinds of materials are typically governed by Library license agreements:

    • Materials you access through the Library's search tools (the Library’s online catalog)
    • Articles from academic journals accessed through the Library
    • E-books available through Library databases
    • Research datasets licensed by the Library
    • Any content accessed through Library database subscriptions
    • Materials that require you to log in with your UNCW credentials
  • Full licenses for many publishers, specifying all the terms and conditions, are available on their websites. If you have questions about whether a particular agreement allows what you want to do, please contact the Open Knowledge & Research Impact Librarian using the links to the left of this page.
  • Here is a list of a few prominent publisher's policies related to AI:

Attribution

In addition to sources linked throughout the text, this guide drew information from the following sources: 

CC License

 

 

 

 

“Copyright and AI” by Sam Winemiller is licensed under Creative Commons Attribution 4.0.