Copyright Guidance for Researchers Using AI
Copyright impacts AI at several levels: training models, ownership of outputs, and the potential for outputs to be infringing. You need to understand your rights and responsibilities in all of these areas.
AI is evolving quickly. We are always happy to discuss current issues regarding ownership of outputs, how your content or data may be reused, and/or how you might use content or data for AI training purposes.
Contact the Open Knowledge Center
Training generative AI models
The state of the law on training AI is uncertain, with ongoing litigation in several jurisdictions. Many journals are licensing their content for use in AI training. You can articulate your preferences when you publish by negotiating your publication agreement.
Publishers may require particular ways of documenting your AI use or may seek to forbid its use. Publishing contracts may also ask you to allow your work to be used or repurposed by AI.
If you have questions about AI terms in a contract, please contact the Open Knowledge Center for support.
Ownership of generative AI output
Generally speaking, if the outputs of AI are eligible for copyright protections they will belong to the creator. But that’s a big “if”. The United States Copyright Office has offered guidance suggesting that copyright does not extend to purely AI-generated material, or material where there is insufficient human control over the expressive elements. While these questions may be determined on a case-by-case basis, the Copyright Office also notes that prompts by themselves do not provide sufficient human control to qualify for copyright protection.
However, human authors are entitled to copyright in their works of authorship that are perceptible in AI-generated outputs, as well as the creative selection, coordination, or arrangement of material in the outputs, or creative modifications of the outputs. In those cases, the human who created the works, rather than the creator of the AI tool, is likely to be the rightsholder, based on the license terms of the tool. Questions? Contact the Open Knowledge Center
Potential for infringement
AI-generated works can be infringing, just like any other materials. The same questions about originality, idea/expression, and creativity apply, as do all exceptions such as fair use. If you are considering using AI-generated materials in your own work, please reach out to the OKC for more information
Pay particular attention when using data that is not your own with AI tools. Even if fair use applies to your own use of data, it may not give you the right to share it with particular AI tools, especially those not approved by NC State's Office of Information Technology (OIT).
Before using any campus data with AI tools, determine the data classification level and review OIT's list of approved AI tools for different data classification levels.
For a helpful primer on AI copyright when it comes to resources accessed through the Libraries, check out "Before you scrape and before you train" from UC Berkeley Libraries.
If you are collaborating with partners outside of the U.S., please contact us to discuss what other legal structures may impact your research.