Generative artificial intelligence (AI) programs, like Dall-E and ChatGPT, seem to be all the rage right now. Tech companies big and small are now racing to come up with the next big thing. Like other forms of AI, generative AI analyzes large amounts of data to identify which patterns will be used to create some output. What makes generative AI different is that its output can be used as content, such as text, images, music and videos. In a world where content is king, generative AI has the potential to transform entire industries, including gaming and esports.
Generative AI may be used to create new maps, characters, storylines and even new video games. Such content is generally protected under U.S. copyright law – which is what allows video game developers and publishers to have more control over their games than owners or leagues have in traditional sports. However, AI-created content may not be entitled to any copyright protection, depending on the level of human involvement. For example, after issuing a copyright registration to Kristina Kashtanova for Zarya of the Dawn, a graphic novel featuring images generated using Midjourney AI, a text-to-image AI, the U.S. Copyright Office sent a letter on February 21, 2023, regarding their decision to cancel the original registration and issue a new one that specifically excluded content created by Midjourney AI, i.e., the images. The new registration only covered Kashtanova’s contributions, which consisted of the text and the selection, coordination and arrangement of the written and visual elements.
The U.S. Copyright Office has long required human authorship for registration and has consistently refused to register works created solely by non-humans. (See §§ 306 and 313.2, Compendium of U.S. Copyright Office Practices (3d ed.)). But, in light of technological advances, is this necessarily the right approach?
In a case before the U.S. District Court for the District of Columbia, Dr. Stephen Thaler, who created Device for the Autonomous Bootstrapping of Unified Sentience, an AI system known to the public as "DABUS," sued the U.S. Copyright Office for refusing to register art created by DABUS because it did not satisfy the human authorship requirement. In the U.S., copyright protection exists in an original work of authorship as soon as it is created, with the rights initially vesting in the author. 17 U.S.C. §§ 102 and 201. Typically, that's going to be the person who created the work; however, under current copyright laws, authors need not be human. That's because "works made for hire" aren't authored by their creators, but rather by those who hired their creators, yielding some non-human authors, like corporations and LLCs.
In dueling motions for summary judgment filed earlier this year, Dr. Thaler argued that AI-generated works were copyrightable, and that, as the owner of the AI, he owns the copyright. The U.S. Copyright Office argued that human authorship is necessary to sustain a copyright claim under the Copyright Act and that the Works Made for Hire Doctrine does not apply in this case because DABUS is not a person, employee or agent. The case is still ongoing, and the motions have yet to be decided.
Perhaps a bigger question when it comes to generative AI is in regard to copyright infringement. Copyrighted works may be used to build or train at least some generative AI. While large-scale web scraping of copyrighted material may raise a whole host of legal and ethical issues, from a copyright perspective, there may be good, fair use arguments for using copyrighted material as input for generative AI if its use is transformative and doesn’t impact the market for the copyrighted material. See Authors Guild v. Google, Inc. 804 F.3d 202 (2d. Cir. 2015).
That said, even if the use of copyrighted material as input may be considered fair use, generative AI could still generate output that constitutes infringement, particularly if that output could be deemed a commercial replacement for the copyrighted work. This is certainly possible where generative AI is already being used to produce works that are "in the style of" specific artists or performers. Companies that deploy generative AI may have terms of service that purport to limit or prohibit use of copyrighted material, but they could still face secondary liability if they encourage and profit from their users' infringement. See MGM Studios, Inc. v. Grokster, Ltd., 545 U.S. 913 (2005). Moreover, generative AI companies may not be afforded the same protections as other online platforms for its users' actions because generative AI may be seen as an "information content provider" under Section 230 rather than an "interactive
Still in its infancy, the possibilities of generative AI seem to be endless. The proliferation of generative AI has raised a lot of interesting questions about authorship and copyrights that will likely change the legal landscape for years to come. Companies using generative AI should be proactive about staying on top of the latest developments in both technology and the law.