November 28, 2023 – ChatGPT and other artificial intelligence (“AI”) programs, such as the image-generating Midjourney, have sparked discussions on generative AI, prompting significant public policy inquiries into the protectability of AI-generated works.
The full extent of generative AI’s impact is yet to be fully realized, presenting all creative domains with an unpredictable future amid rapid transformations. Despite this uncertainty, the U.S. Copyright Office has initiated steps by launching a formal AI initiative, albeit without establishing concrete rules. This initiative has set the stage for a pivotal debate: What criteria must be met to fulfill the Copyright Office’s requirement of human authorship to grant copyright protection to AI-assisted creative content?
Current Status of the Debate and Its Relevance to Creative Fields
Many artists view AI as “plagiarism machines,” especially concerning the unauthorized use of their copyrighted material as references by AI applications. Generative AI generates output by leveraging learned patterns from existing works. While some argue that this process mirrors traditional artistic inspiration, recent litigations suggest otherwise. Plaintiffs have contended that generative AI inherently relies on existing works to train its algorithms. Recent disputes, such as the resolved conflicts involving the Writers Guild of America (“WGA”) and the Screen Actors Guild (“SAG-AFTRA”), underscore the crucial need to regulate the use of generative AI in creative sectors.
During negotiations between the WGA and the Alliance of Motion Picture and Television Producers (“AMPTP”), the WGA pushed for restrictions on AI’s involvement in creating or revising content and using material from the Minimum Basic Agreement to train AI models.
Conversely, entities like AMPTP recognize the value of rapidly evolving AI technologies. For instance, Disney has explored AI applications since June 2020 to develop characters for its Star Wars franchise. The resolution between the parties incorporated several demands, including defining AI-generated content as not constituting “source material” and granting the WGA the right to claim if a writer’s material was used to train an AI model. Refer to “Why AI Is the Most Important Issue in the Writers’ Strike,” The Writing Cooperative, May 25, 2023, for further insights.
The ongoing debate surrounding these disputes might become obsolete as new regulations emerge from the Copyright Office or potential federal legislation. However, policymakers must balance regulating AI advancements with fostering innovation to maintain a competitive edge globally.
Insights from Copyright Office Guidance and Rulings
Exclusion of Copyright Protection for AI-Generated Works Lacking Human Authorship
The Copyright Office has been pivotal in adjudicating AI-related cases and providing regulatory guidance. In 2022, Stephen Thaler’s attempt to copyright a two-dimensional AI-generated image named “A Recent Entrance to Paradise” using his AI tool, the “Creativity Machine,” was rebuffed by the Copyright Office due to the absence of “human authorship essential for copyright eligibility.”
In response, Thaler contested the decision, asserting ownership of the work as a product made for hire by the AI tool. On March 16, 2023, the Copyright Office issued a “Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence” to clarify its practices concerning works incorporating AI-generated content.
Media coverage largely focused on the unprotectability of AI-created works. However, overlooked was the potential for copyright protection when a human collaborates with generative AI under specific conditions.
The March Statement of Policy addressed two critical queries raised by Thaler and others utilizing generative AI: whether a generative AI can be deemed an author and if a work produced by AI can be copyrighted as a work made for hire by the individual providing prompts.
Regrettably for AI proponents, both queries received a definitive “No” from the Copyright Office, emphasizing that copyright safeguards solely materials stemming from human creativity. Consequently, when AI solely responds to prompts, the resultant work lacks copyright eligibility due to the user’s limited creative influence over the final output.
Determining the Threshold of Human Involvement for Copyright Protection
The March Statement of Policy also tackled the registration of works comprising both human-authored and AI-generated content. It stipulated that if human intervention creatively manipulates AI-generated content, the resulting work becomes eligible for copyright protection. Therefore, while the AI-generated material remains unprotected, the altered work qualifies for copyright.
Furthermore, the Statement hinted at the possibility of copyright protection for works where user-prompts lead to a more anticipatory expressive outcome. This scenario raises questions about the necessity of a complete writers’ room in the future. Could generative AI draft initial versions for subsequent human editing?
The case of Maria Kashtanova and her graphic novel “Zaraya of the Dawn” sheds light on the collaborative process between humans and AI. A memo from the Copyright Office on Feb. 21, 2023, highlighted how substantial human editing of AI-generated content satisfies the human authorship prerequisite for copyright protection.
While headlines sensationalized the denial of copyright claims for AI-generated images, the ruling acknowledged the copyrightability of Kashtanova’s graphic novel as a whole:
The Office affirms that Ms. Kashtanova’s selection, coordination, and arrangement of the text and visual elements in the Work demonstrate human authorship, warranting copyright protection.
In another instance, an April 20, 2023, letter from the Copyright Office to The Mechanical Licensing Collective reiterated the need for human authorship in music creation. While purely AI-produced musical works lack copyright eligibility, human intervention in selecting or arranging music creatively renders the resulting work copyrightable.
Future Implications for Creatives and Copyright
The prevailing narrative depicting generative AI as a threat to human creativity across various sectors like journalism, media, art, and entertainment may seem exaggerated in the short term. Nevertheless, the integration of generative AI is poised to alter creative processes significantly. A conceivable scenario involves AI generating initial drafts subsequently refined by humans. Conversely, human-authored works could undergo AI editing or manipulation while retaining copyright protection.
Despite gaining clarity on the requisite human authorship level for copyright protection in AI-augmented creative works, challenges persist. For instance, how should content creators be compensated if their works train AI models? Class-action suits filed by Sarah Silverman and others against OpenAI and Meta underscore this issue, along with an open letter from over 8,000 authors demanding compensation for unauthorized use of their works in AI training.
In correspondence with Stephanie Weiner, Acting Chief Counsel of the National Telecommunications and Information Administration on June 12, 2023, the Copyright Office disclosed plans to address whether using existing works to train AI models constitutes fair use. A forthcoming notice of inquiry will solicit public feedback on this matter. Notably, countries like Japan and Israel have already implemented fair use exceptions for model training to foster innovation.
The outcomes of ongoing debates will profoundly impact creativity, ownership, and control in the creative landscape. Businesses, technology firms, legal entities, and creative professionals must stay abreast of evolving trends to address these issues effectively in their practices and agreements.
