Why AI-Generated Works Should be Protected
Both the Constitution and the Copyright Act's history and objectives support protecting AI-generated works. Such protection would encourage the development and use of creative AI technologies, promoting the distribution of socially beneficial works, aligning with copyright law's aims. Without such protection, some AI-generated works might not be developed or shared, as significant time and financial investments are often necessary for their creation and dissemination, similar to human-created works.
While it could be possible to designate a natural person under non-traditional criteria as the author of an AI-generated work, such as an AI’s programmer or a user providing a prompt to an AI, it may be most appropriate to list an AI as an author for an AI-generated work. That is because it is factually accurate and prevents someone from taking credit for work they have not done. Of course, protecting AI-generated works and listing AI as an author for copyright purposes is not about granting legal rights to machines, as AI is not a legal entity and cannot possess rights or obligations. Instead, protecting AI-generated works would foster transparency, appropriate allocation of rights, and informed policymaking.
If AI-generated works are to be eligible for copyright protection, this raises the question of who should be the owner of such copyright. Obvious candidates include an AI’s programmer, user, or owner. This article suggests that making an AI’s owner the owner of any works it generates is most consistent with general property law principles, such that the owner of property is the owner of property made by their property, whether that is fruit from a tree, a physical painting made by a 3D printer, or a digital painting made by a generative AI. Ultimately, the specific identity of the default owner is less crucial than the establishment of clear property rights. Where different parties are involved—users, programmers, owners—contracts can dictate optimal arrangements, especially where copyright is a key creation or collaboration incentive.
By contrast, some commentators believe that AI-created works should not be copyrighted, either for moral reasons—arguing that AI creations do not merit protection—or for economic reasons—suggesting that AI does not require financial motivation to create, or that protection is unnecessary for other reasons, or that the costs of protection outweigh the benefits. In the United States, the Copyright Office has had a “Human Authorship Requirement” policy since at least 1973, disallowing copyright registration for AI-generated works. While the Copyright Act does not explicitly state that an author must be human or that human creativity is necessary for copyright, legal precedents have historically viewed creativity through a human-centric lens. However, it is worth noting that other forms of artificial authors—like corporations and nations—have been considered authors under the Copyright Act for over a century.
Protection of AI-Generated Works in the United States
Until recently, no case in the United States had directly addressed the copyrightability of AI-generated works. Thaler v. Perlmutter, part of the Artificial Inventor Project led by Ryan Abbott, coauthor of this Article, has sought to challenge the Copyright Office’s Human Authorship Requirement. This project, through a series of pro bono legal test cases, aims to secure intellectual property rights for AI-generated works in the absence of a traditional human inventor or author. Its purpose is to spark discussion on the societal, economic, and legal impacts of cutting-edge technologies like AI and to provide guidance on the protectability of AI-generated outputs.
In 2019, the Copyright Office declined to register "A Recent Entrance to Paradise," a 2D artwork generated by an AI system owned by Stephen Thaler. Dr. Thaler, as the owner, user, and developer of the Creativity Machine, attempted to register the artwork but faced rejection due to the absence of traditional human authorship. After two reconsideration requests, the final decision on February 14, 2022, upheld the rejection, affirming human authorship as a prerequisite for copyright protection in the U.S. Subsequently, Thaler filed a lawsuit in June 2022 against the Copyright Office, seeking to compel the registration of the artwork with the AI listed as the author and Thaler as the copyright owner.
Other applicants tested the system by trying to register works created with text-to-image generators. One notable case involved artist Kristina Kashtanova, who registered a comic book titled "Zarya of the Dawn," with images produced using the generative AI system Midjourney. Initially accepted on September 15, 2022, the registration faced reconsideration after Kashtanova publicized the AI's role on social media. The USCO decided on February 21, 2023, to maintain copyright for Kashtanova's text, selection, coordination, and arrangement, but not for the Midjourney-generated images. This led to the cancellation of the original certificate and the issuance of a new one excluding the AI-generated content.
Following this decision, on March 16, 2023, the USCO released guidance on works containing AI-generated material. The USCO has clarified that the duty to disclose AI-generated elements depends on whether the AI-created content in a work is significant or de minimus. The criterion used to determine significance is whether the content would be copyrightable if created by a human.
Subsequently, in Thaler v. Perlmutter, the court ruled in favor of the Copyright Office on August 18, 2023, reiterating that human authorship remains a fundamental requirement for copyright. Judge Beryl A. Howell, acknowledging copyright law's adaptability to new technologies, nevertheless upheld the principle that human creativity is central to copyright eligibility. The case is now under appeal to the Court of Appeals for the District of Columbia Circuit.
Protection of AI-Generated Works Internationally
By contrast to the U.S., some jurisdictions provide copyright protection for AI-generated works. The United Kingdom, for instance, amended its copyright law in 1988 in part to explicitly provide copyright for AI-generated works, albeit with a shorter term of protection, namely 50 years from the date of creation vs. 70 years plus the life of an author. The person who undertakes to have the work created is legally deemed or fictionalized to be the author. Other jurisdictions, like India, South Africa, Ireland, and New Zealand have similar statutory provisions.
Internationally, most jurisdictions lack definitive rules on the protectability of AI-generated works, and some are actively reevaluating their existing policies. International copyright agreements neither explicitly allow nor forbid protections for AI-generated works. Yet, under such agreements, if a country like the United States offers such protection, it must extend this protection globally, even to works created in countries that might not offer similar protection. This scenario leads to a disparity where consumers in certain countries, but not others, can freely use and copy AI-generated works from any location. This could be viewed as unfair, allowing some countries to benefit from the creative output of others without offering similar protections. Conversely, countries that primarily import AI-generated works might benefit from not providing such protections.
Technological advancements often prompt copyright law reassessment. With AI now significantly contributing to creative works that hold consumer value, its ever-improving capabilities and the challenges in attributing AI outputs directly to human creators call for a comprehensive reevaluation of what copyright rules will maximize social benefits.
