Although artificial intelligence (AI) has been assimilated into most of our lives in one way or another for many years, the recent increase in visibility of “generative AI” systems and other AI platforms, such as Open Art, Stable Diffusion, Deep Dream, ChatGPT, and DALL-E, has catapulted (or re-catapulted) AI into the forefront of public curiosity and legal commentary. Reactions ranging from excitement, trepidation, and awe are churning amidst questions surrounding the potential ramifications that such technology may have on education and many creative and professional fields. From an intellectual property law perspective, this has revived longstanding debates and is likely to spark fresh questions around the ownership, protection, and enforcement of intellectual property rights, as relevant technology continues to develop and spread in application.
Turning specifically to copyright, the form of intellectual property protecting original works of authorship (i.e., literary, musical, dramatic, and artistic works), this is certainly not the first time that questions regarding the ability to protect and own the work of non-human contributors has come before the US Copyright Office or courts.
Both the Copyright Office and courts have taken the position that copyright can protect only material or content created by a human being. The term “author,” used in the Constitution as well as the Copyright Act, is found to exclude “non-humans.” In its “Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence” dated March 16, 2023, the Copyright Office refers back to the Supreme Court’s leading case on authorship, Burrow-Giles Lithographic Co. vs. Sarony, which held that photographs are permitted to be subject to copyright, despite the argument that they are not created by a human but by a camera, focusing instead on the extent to which photographs are representative of the author’s “original intellectual conceptions.” The Copyright Office’s guidance also briefly touched on the Ninth Circuit decision regarding the now-famous “monkey selfie,” and that the wording of the Copyright Act refers to terms that “imply humanity and necessarily excludes animals.”
Ultimately, the question of the extent to which the use of AI or other tools may impact authorship is a fact-specific inquiry, perhaps roughly summarized as an evaluation of the determination and execution of the traditional elements of authorship (such as the author’s original mental conception and exercise of creative control) and whether such elements are carried out by a human or machine/non-human system. For example, where many generative AI systems effectively produce content based on prompts input by a human, such prompts are likened to instructions given to an artist in the commissioning of a work. In each case, the AI system and commissioned artist are likely the determiners of how the instructions are actually implemented into the resulting material and the person giving the prompt or instructions is not exercising creative control.
That being said, the mere inclusion of some AI-generated content does not preclude copyright protection as to a work in its entirety. In some cases, there may be sufficient human authorship to support a copyright claim; however, copyright protection is limited to the human-authored aspects and any protectability of such aspects does not impact the copyright status of AI-generated material. In applying to register works created using AI technology, applicants may claim copyright protection as to their own contributions to such work. Any AI-generated material included in or incorporated into such work should be disclosed and, likely, expressly excluded from the application and claim for protection. Applicants must carefully assess the amount and scope of human authorship in a work, as failure to properly and accurately claim authorship may place a registration and the ability to take action against others for infringement at risk.
Outside of seeking copyright protection, those looking to use AI in preparing or developing materials should exercise caution in turning to generative AI. Aside from the potential for uncertainty as to the ability to exclusively or even sufficiently own, control, or protect content generated by an AI system, a plethora of issues may arise in connection with the content that was input into or otherwise used by an AI system to generate the content produced, including as to the accuracy, comprehensiveness, and third party ownership of the same.
As artists are finding that their works are being “scraped” from the internet or possibly targeted more specifically to be copied or imitated by AI, questions regarding the legality and ethics of how generative AI systems are “trained” continue to be raised. Earlier this year, Getty Images filed suit against Stability AI, alleging that the tech company behind Stable Diffusion committed copyright infringement in its copying and processing of images owned or represented by Getty Images without an appropriate license. In response to such claims, many companies developing the types of AI technology in question are likely to assert “fair use” under US copyright law.
The determination of fair use is highly fact-specific and weighs multiple factors, including the purpose and nature of the use (including use for educational versus commercial purposes), degree of transformation, and impact on the market for the original work. How persuasive the argument for fair use is, if made, could differ significantly between the process of training AI or machine learning programs and the process of producing output content, where the input copyrighted works may remain discernable. Even so, AI companies may take the position that uses of the copyrighted works are sufficiently transformative and represent new creations, which would need to be considered in balance with the other fair use factors.
As AI technology develops and integrates in ways that may or may not be obvious, we can expect that it will have significant implications and generate new considerations across all manner of industries. It remains to be seen how copyright and other areas of law may evolve alongside the continued growth of generative AI and machine learning technology.