Intellectual property law across the globe is reckoning with the implications of generative AI for creators of copyrighted works and the artists, programmers, and scholars who would use those works for generative AI outputs. There are many questions implicated, including the copyrightability, authorship and ownership of generative AI outputs; the merits of a fair use defense when using copyrighted works to train AI tools; and what regulations, if any, should be imposed in response to this dramatic shift in technology and society. At this time, there is not a clear answer for many of these questions.
Most copyright regimes have certain requirements that must be satisfied before protecting a work under copyright, patent, trademark, or other forms of intellectual property law. These requirements may be based on the subject matter of the output; the originality and creativity vested in the work; or a standard of fixation. (In fact, if a prompt given to a generative AI tool satisfies these requirements, it may be eligible for copyright protection itself, even if the resulting output is not).
These requirements can be found in the text of §102 of the US Copyright Act (17 USC) and in Chapter I of the Copyright Law of the People’s Republic of China (中华人民共和国著作权法)
Beyond this, most copyright regimes have requirements with regards to the authorship of a work in question before conferring copyright protection to a work.
The answer to questions about ownership and authorship of AI works under patent, trademark, and copyright laws will vary across jurisdictions and whether the work is “AI-generated” or “AI-assisted”. The World Intellectual Property Organization (WIPO) Secretariat offers definitions in paragraph 12 of the Revised Statement on Intellectual Property Policy and Artificial Intelligence (WIPO/IP/AI/2/GE/20/1 Rev.):
“AI-generated” and “generated autonomously by AI” are terms that are used interchangeably and refer to the generation of an output by AI without human intervention. In this scenario, AI can change its behavior during operation to respond to unanticipated information or events. This is to be distinguished from “AI-assisted” outputs that are generated with material human intervention and/or direction.
When considering if your AI-generated or AI-assisted output is copyrightable, expert legal advice should be sought.
In the United States, authors of a copyrighted work must be human: the 9th Court of Appeals affirmed in Naruto v. David Slater et al. (2016) that non-human animals lack statutory standing for protection under the Copyright Act, and the US Copyright Office has refused to register AI-generated works produced autonomously without human involvement. However, in Copyright Registration Guidance for Works Containing AI-Generated Material, the Office distinguishes between works that “are the result of “mechanical reproduction” instead of an author’s “own original mental conception, to which [the author] gave visible form.” (Quoting Burrow-Giles v. Sarony).
The question of protecting AI-generated works has already arisen in the context of Chinese copyright law, as well. The Cyberspace Administration of China (CAC, 国家互联网信息办公室) is drafting measures for the management and regulation of generative AI systems, open for public comment until 10 May 2023.
In 2019, the Nanshan District People’s Court in Shenzhen drew a distinction between “AI-generated” and “AI-assisted” outputs in Shenzhen Tencent v. Shanghai Yingxun (you can read the full text of the court’s decision here, though you’ll need to register with a phone number first). The Court granted copyright protection to the works in question, determining that the content generated were written works (a copyrightable subject matter) and that the creative team was sufficiently involved in the input, selection, arrangement, and other intellectual activities that the works were original and creative enough to be ultimately the product of human authorship.
When using emerging tools for generative AI projects, researchers, artists, programmers, and others should consider the expertise of legal counsel in guiding their use of training data ethically, as well as protecting the outputs of those tools.
Although scraping content to create datasets may often be considered an example of "fair use" in US copyright law, that does not mean that all generative AI projects will fall within the ambit of this exception. Claiming a fair use defense for any generative AI project will need to be assessed on the facts of that specific instance using the four-factor test.
Data and materials used to train generative AI tools may themselves be copyrighted and protected by both property and personality (or "moral") rights. In recent years, owners of copyrighted works have demonstrated a willingness to challenge the unauthorized use of their intellectual properties.
In 2021 the estate of artist Pablo Picasso invoked the author’s moral right to disclose a work to demand that his works not be used without permission for generative AI artworks.
In 2023, multiple lawsuits were brought against unauthorized uses of artworks for training generative AI tools and producing new visual works. In the United Kingdom, Getty Images filed a complaint against Stability AI with the High Court of Justice in London. Getty argues that they offer licenses for uses such as these, and that when Stability AI ignored these viable licensing options Getty’s copyright was infringed. In the United States, a class action lawsuit was filed on behalf of three artists against Stability AI and others. Counsel for the plaintiffs does not mince words, likening Stability AI to “a parasite that, if allowed to proliferate, will cause irreparable harm to artists, now and in the future.” You can view the complaint & exhibits from the plaintiffs here. Thus, it is not clear-cut that using existing copyrighted works to train generative AI tools will be covered by a fair use (or similar) defense.
For Research Support Contact
Caitlin MacKenzie Mannion
Head of Reference & Instruction Services and Librarian for the Arts & Humanities
Email: shanghai.librarian@nyu.edu
Book an appointment: https://nyu-shanghai.libcal.com/appointments/CMKM
The library is unable to provide legal advice or guidance. If you have specific questions about your project’s compliance with local IP Law, please seek official legal counsel.
NYU Shanghai community members may also contact the University’s office of Legal Compliance.