Close
Close
This guide © 2024 by UCL - Library Skills is licensed under CC BY-NC-SA 4.0 
The prompts you have created yourself and provide as inputs to the tool may be protected by copyright, as long as they are original, i.e. your own intellectual creation. Examples include a series of short prompts you provide as instructions to create and refine an image; and articles, thesis extracts or essays you have written yourself.
Even if the prompt is your own intellectual property you should be aware that prompts you provide may be used to train the tool and reproduced in future outputs. Whether and how different tools do this varies and should be set out in their terms and conditions. They may state that they will not use your data, provide this as an opt-in or have a default licence for reusing the data, offering an opt-out. Licences to the tool to reuse your inputs and outputs can be very broad. Read the terms and conditions and if necessary, opt out of your data being reused, particularly if there are data protection concerns or confidentiality agreements. Please note that inputting commercially sensitive information that could, for example, lead to a patent would be considered as disclosure and restrict further exploitation. UCL’s IP policy also advises on this.
Are GenAI outputs protected by copyright in the first place? The approach to this question varies across countries and it is likely to vary on a case by case basis, too, depending on whether the output is deemed to be original, reflecting the its author’s creative choices.
UK copyright law allows for the possibility that GenAI outputs are protected by copyright, but this has not been tested in court. Section 9(3) of the Copyright Designs and Patents Act states that ‘in the case of a literary, dramatic, musical or artistic work which is computer-generated, the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken’. It is not specified whether this would be the user who provided the prompts or the developer of the tool; this is something that may be determined in the terms and conditions of the tool.
It can also be argued that a GenAI output is not original in the sense of being a person’s own intellectual creation, and therefore is not protected by copyright. This is the approach taken so far in the US. One example is the graphic novel Zarya of the Dawn created by Kristina Kashtanova, which includes a story and characters created by the author and images created using the GenAI tool Midjourney. When addressing authorship and protection of the work, the court decided that the individual images are not protected by copyright as they were not created by a human. While prompts were provided by Kashtanova to produce the images, these were perceived as general instructions rather than creative input. For more information on this case see the IPKat website.
Courts in China have taken a different approach. In Li vs Liu, the court ruled that an image created by using about 150 prompts, negative prompts and modifications is protected by copyright and that the author is the person who provided the prompts, as the result reflected their own creative choices. Reuse of the image by a third party was therefore deemed to be infringement and damages were paid to the author. This case is discussed on the TechnoLlama website.
Emerging cases highlight that originality, copyright protection and authorship in the context of GenAI are approached differently across jurisdictions.
Generative AI has immense potential to support research when used in an open, critical, ethical, transparent and informed way. Training datasets that consist of high quality research should help increase the accuracy of the models and reduce biases. For a more extensive discussion, please see the Knowledge Rights 21 Principles on Artificial Intelligence, Science and Research.
Scholarly research that is published open access or shared in open access and research data repositories could be used to train AI under the licensing terms that apply. If text and data mining or other copyright exceptions apply, subscription articles, monographs and book chapters may also be used as training data as long as the provider has lawful access. This raises attribution issues and copyright issues, for example if the outputs reproduce substantial content from the originals or create derivatives.
A number of publishers now have, or are negotiating, agreements with AI providers licensing the reuse of their publications. While it has been assumed that publishers, traditionally being the copyright owners of journal articles, can license these works without the consent of their authors, it can also be argued that such use is not covered by previous agreements and permission should be sought separately.
Publishers may now ask you to agree to your article, monograph or book chapter to be used as part of training datasets. Such agreements should address:
You should be able to ask your editor for more information before you make a decision.
A comprehensive discussion of issues related to GenAI and academic publishing is available on the Kluwer copyright blog:
Trapova, Alina (2024). 'Report on a roundtable on academic publishing and genAI deals – GenAI and copyright series at the Institute of Brand and Innovation Law', Kluwer Copyright blog, 18 December. Available at: https://copyrightblog.kluweriplaw.com/2024/12/18/report-on-a-roundtable-on-academic-publishing-and-genai-deals-genai-and-copyright-series-at-the-institute-of-brand-and-innovation-law/ (Accessed 18 December 2024).
This section addresses issues arising around reuse of materials you created yourself, e.g. prompts you wrote or research articles that could be used as training data. Copyright protection and copyright ownership of these materials are also addressed.
