ChatGPT: Who Owns the Content Generated?
In ChatGPT: Legal Counsel of the Future, we described the Chat Generative Pre-Trained Transformer (“ChatGPT”), how it works, and whether or not ChatGPT can provide legal advice. The language learning model has many potential uses, such as debugging code or writing essays.¹ The use raises questions centered around intellectual property. For example, who owns the content ChatGPT generates? Could one be infringing existing intellectual property rights when using the model? Are there confidentiality issues? Can I use ChatGPT to create intellectual property? If so, how? This article explores these intellectual property issues and discusses using ChatGPT to develop intellectual property.
Ownership
When discussing ownership, it is worth mentioning what ChatGPT tells us about ownership.
Interestingly, the language model states that the responsibility for the content ChatGPT generates lies solely with the user but that the use of the model is subject to the terms and conditions set by OpenAI, LLC (“OpenAI”). Section 3 of OpenAI’s Terms of Use deals with content.² Content is defined as input that the user provides to an OpenAI service, such as ChatGPT, and the output generated and returned by ChatGPT based on the input. ChatGPT states that the user owns all input to the extent applicable law permits. Additionally, OpenAI assigns all its right, title, and interest in and to the output to the user.
It should be noted that OpenAI only assigns all “its” right, title, and interest in and to the output to the user. However, if OpenAI does not initially own the rights, it cannot assign them. Another related issue is that other ChatGPT users can narrow the assignment scope. OpenAI’s Terms of Use state that due to the nature of machine learning, many users may receive identical or similar outputs from ChatGPT.³ This leads to another set of issues, including determining original ownership rights (if available), the potential destruction of ownership rights over time, and murky usage rights in the future.
The ownership question is challenging because the generated output is created by artificial intelligence (“AI”). Determining ownership of AI-generated content is particularly difficult where patents and copyrights are concerned. In the US, a human must be the inventor of a patentable invention or the author of a copyrightable work.⁴ An attempt to have an AI machine recognized as the sole inventor has been unsuccessful.⁵ However, the question of whether inventions made with the assistance of AI are eligible for patent protection has not yet been dealt with. In a copyright case dealing with the authorship question, it was held that a monkey that took photos of himself had no rights to the photographs.⁶
However, last month, the US Copyright Office issued guidance regarding copyright registration of works containing material generated by AI.⁷ To be able to register a work incorporating AI, the applicant must disclose the inclusion of AI-generated content in a work submitted for registration and provide a brief explanation of the human author’s contributions to the work.⁸ The guidance indicates that works containing AI elements can include sufficient human authorship to meet the standard for copyright protection. For example, where the human selects or arranges AI-generated material in a sufficiently creative way that the resulting work as a whole constitutes an original work of authorship or where the human modifies material originally generated by AI technology to such a degree the modifications meet the standard for copyright protection.⁹ However, it seems that works created by language models such as ChatGPT, absent sufficient modification by a human, will not be protected by copyright:
“Based on the Office’s understanding of the generative AI technologies currently available, users do not exercise ultimate creative control over how such systems interpret prompts and generate material. Instead, these prompts function more like instructions to a commissioned artist—they identify what the prompter wishes to have depicted, but the machine determines how those instructions are implemented in its output … When an AI technology determines the expressive elements of its output, the generated material is not the product of human authorship. As a result, that material is not protected by copyright and must be disclaimed in a registration application”
Id., at 16192.
As far as current copyright law is concerned, it seems likely that an AI-generated work created by a language model such as ChatGPT would either become part of the public domain at the moment of its creation or potentially be treated as a derivative work of the materials used to train the AI tool.¹¹ Due to the sheer volume of text-based training data ChatGPT was provided with, it will likely often be extremely difficult to determine which dataset was ultimately used to generate a particular output. In summary, although ChatGPT is a great tool, the user will likely not be considered the author or inventor of the generated output.
Infringement of Intellectual Property Rights
The first question that comes to mind is, does using output generated by the model potentially automatically infringe the intellectual property rights of another? The answer to this question is highly relevant to the user, as the OpenAI Terms of Use make the user “responsible” for the Output.¹²
How copyright and patent infringement will be handled still needs to be determined. Where copyright is concerned, one issue will be determining the infringer. It seems possible that the user would be jointly liable for infringing another’s reproduction right or right to create derivative works. For example, a large language model trained using copyrighted material could excessively copy another’s work or create a derivative work, for example, a sequel to an existing written work, when generating a response to a prompt provided by a user. It should be noted that a user of ChatGPT is required to indemnify OpenAI for any potential claims brought against the company.¹³ Another problem exists when two parties provide prompts that generate identical or similar works. For example, one party cannot bring a successful copyright infringement claim against the other since each created the work independently.
AI could also potentially be used in a way that infringes a patent, to the extent that the AI practiced an infringing process or produced an infringing article. Written content such as essays, poetry, works of fiction, or answers to math problems would be unlikely to infringe a patent as such materials would be considered “printed matter” or abstract subject matter beyond the scope of typical patent protection. However, it is possible to imagine that AI-generated computer code could infringe a patent claim protecting a software or computer-implemented invention. Again, it must be determined to what extent the user and language learning model owner and developers of the AI, are liable for infringement. Although it is currently unclear how these and other infringement issues will be dealt with, both patent and copyright infringement issues will undoubtedly be litigated, providing us with clarity on infringement pitfalls when using ChatGPT.
Confidentiality
When using a language model such as ChatGPT, it’s essential to consider whether the information the user discloses to it is protected. The OpenAI Terms of Use merely provide a unilateral confidentiality provision in favor of OpenAI.¹⁴ Any confidential information provided could be used and disseminated freely by OpenAI, subject to applicable law. The FAQs make it quite clear that a user should not expect confidentiality. “As part of our commitment to safe and responsible AI, we review conversations to improve our systems and to ensure the content complies with our policies and safety requirements.”¹⁵ “Your conversations may be reviewed by our AI trainers to improve our systems.”¹⁶ “[W]e are not able to delete specific prompts from your history. Please don’t share any sensitive information in your conversations.”¹⁷ Therefore, a user should avoid inputting confidential information into the language model.
Using ChatGPT to Create Intellectual Property
Although the use of ChatGPT raises ownership and infringement issues, the model can still be a valuable tool for creating intellectual property. For example, ChatGPT has been used, with varying degrees of success, to write the background section of a patent, draft patent claims, perform a patent infringement search, create patent infringement charts, and create a brand.¹⁸ When using the language model, the consensus is that the individual user should know that the model is often wrong and that generated content should always be verified for accuracy and adjusted as needed. Further, it is generally not advisable to use ChatGPT to draft patent claims and specifications, as a user risks making their invention publicly available before filing the patent application.¹⁹ However, it can provide a user with ideas, help a user while searching for infringement and explain things to them in a simple, understandable way. As stated by Job.com Founder & CVO, Arran Stewart, in a recent talk on HR Tech: The Transformational Power in Recruiting, “As a platform that is going to have various data inputs from so much information because everybody has jumped on ChatGPT, it will have a huge amount of sophistication, but you will have to refine that sophistication for [your specific] purposes.”
In summary, ChatGPT is a remarkable piece of innovation. Its use raises several legal questions, particularly concerning intellectual property ownership, infringement of intellectual property rights, and confidentiality of inputted information. However, AI, such as ChatGPT, can inspire innovation and ideas and enforce existing intellectual property rights. In a decade from now, this technology will significantly evolve and improve, and the treatment of AI may substantially change.
1. OpenAI, Introducing ChatGPT, OpenAI (Nov. 30, 2022), https://openai.com/blog/chatgpt; BBC, Universities warn against using ChatGPT for assignments, bbc.com (Feb. 28, 2022), https://www.bbc.com/news/uk-england-bristol-64785020.
2. OpenAI, Terms of Use, Section 3(a),OpenAI (Mar. 14, 2023), https://openai.com/policies/terms-of-use.
3. Id., at Section 3(b).
4. 35 U.S.C. §100(f); Beech Aircraft Corp. v. EDO Corp., 990 F.2d 1237, 1248 (Fed. Cir. 1993).
5. Thaler v. Hirshfield, 558 F.Supp.3d 238 (E.D. Va. 2021).
6. Naruto v. Slater, 888 F.3d 418 (9th. Cir. 2018).
7. Works Containing Material Generated by Artificial Intelligence, 88 Fed. Reg. 16190-01 (Mar. 16, 2023).
8. Id., at 16193.
9. Id., at 16192-16193.
10. Id., at 16192.
11. Joe McKendrick, Who Ultimately Owns Content Generated By ChatGPT And Other AI Platforms?, FORBES (Dec. 21, 2022, 12:59 PM), https://www.forbes.com/sites/joemckendrick/2022/12/21/who-ultimately-owns-content-generated-by-chatgpt-and-other-ai-platforms/?sh=527b63f65423.
12. Supra, note 2.
13. Id., at Section 7(a).
14. Supra, note 2, at Section 5.
15. OpenAI, ChatGPT General FAQ, Question 5, OPENAI, https://help.openai.com/en/articles/6783457-chatgpt-general-faq (last visited Apr. 4, 2023).
16. Id., at Question 6.
17. Id., at Question 8.
18. Ryan Phelan and Matthew Carey, ChatGPT and Intellectual Property (IP) related Topics (Mar. 27, 2023), PATENTNEXT, https://www.patentnext.com/2023/03/chatgpt-and-intellectual-property-ip-related-topics; Sonam Singh, Infringement Search: AI chatGPT is confident but horribly wrong, HAVINGIP, https://havingip.com/patent-infringement-search-ai-chatgpt (last visited Apr. 4, 2023); Jack Appleby, Can ChatGPT build an entire brand?, FUTURE SOCIAL (Mar. 1, 2023), https://www.marketingbrew.com/future-social/stories/2023/03/01/can-chat-gpt-build-an-entire-brand.
19. 35 U.S.C. § 102(a), (b).