Recent advancements in generative artificial intelligence (GenAI) have significantly influenced how authors work. One potential explanation for this trend is that GenAI may help individuals explore their creativity without being impaired by a general lack of literary knowledge or experience. Consequently, AI-assisted writing platforms, such as Jasper or Sudowrite, have become increasingly popular and commonly used by writers.
Unsurprisingly, this trend has increasingly alarmed authors. Writers are worried about the ethical implications of writing with AI. Some refuse to use AI altogether, claiming that such usage will reduce writing to mere means of communication instead of creative artwork. For example, National Novel Writing Month (NaNoWriMo), a non-profit encouraging creative writing, was boycott after it advocated for liberal standard on use of AI in fictional writings. Beyond these debates, the integration of AI into creative processes has raised complex copyright issues under U.S. law, particularly with rights to authorship, authority, and ownership as humans co-author with AI.
Per U.S. copyright laws, human authorship is a key requirement to qualify for any copyright protection. Trade-Mark Cases have established that U.S. copyright laws only protect “the fruits of intellectual labor . . . founded in the creative powers of the mind.” Burrow-Giles Lithographic Co. v. Sarony later clarified and reaffirmed that the U.S. Copyright Office can refuse to register a copyright claim if a work was not done by a human.
Lower court examples fill in the details. In 2018, the Ninth Circuit decided Naruto v. Slater wherein the Court held that although a monkey might even meet the threshold for Article III standing, a selfie taken by the monkey does not meet the firm requirement of human authorship; the court ultimately dismissed the monkey’s claim to copyright protection. Later in a 2023 case, Thaler v. Perlmutter, the District Court for the District of Columbia held that anything purely generated by AI, without substantial amounts of human creative contribution, is ineligible for federal copyright protections.
The U.S. Copyright Office has released a two-part report in response to the shifting landscape of generative AI, which explores the intersection of copyright law and artificial intelligence. Part one examines issues related to digital replicas, while part two focuses specifically on questions surrounding copyrightability. The report explicitly stated that works fully generated by AI without meaningful human creative input are ineligible for copyright protection and will remain in the public domain because they do not meet the statutory definition of authorship. The report does acknowledge, however, that works created with significant human involvement or creativity in conjunction and collaboration with AI-generated materials may still qualify for copyright protection.
The report specifically categorizes human contributions to AI-generated outputs into three distinct types: “(1) prompts that instruct an AI system to generate an output; (2) expressive inputs that can be perceived in AI-generated outputs; and (3) modifications or arrangements of AI-generated outputs.” The report emphasized for the first type that merely providing basic or generic prompts, even if carefully worded, generally fails to amount to sufficient creative authorship. The requisite threshold of creativity must be greater than mere instruction or standard prompts that do not reflect substantial original expression. As for the second type, expressive inputs entail creative direction and/or expressive elements deliberately incorporated by the human author into the AI-generated output. When such expressive inputs are clearly identifiable and contribute significantly to the final output, the final product may be eligible for copyright protection. Finally, the third type has the strongest claim in favor of copyright protection. The third type is categorized and recognized as when a human author significantly modifies, edits, assembles, or creatively enhances AI-generated materials.
The U.S. Copyright Office has released additional registration guidelines for registering works containing work generated by artificial intelligence. Only human authors may be listed as or considered authors; AI tools cannot be given authorial credit. The Copyright Office also specifies that registrants must provide a brief statement explaining the authorship of the content.
Despite the broad definition, AI-generated works still need to clear a high-bar for copyright protections. In 2022, Jason Allen took first place in the digital category at the Colorado State Fair with an AI-generated image. Even with the award recognition, the U.S. Copyright Office ultimately rejected his request for copyright registration. Allen argued that his prompt input of more than 600 words constitutes human authorship. In response, the Office disagreed, stating that the prompt does not fully control the AI algorithm used. Therefore, Allen cannot be considered as the author of the image. Nevertheless, the Office left it open the possibility of registration if Allen can provide more information on edits he made with Photoshop.
Despite the recent details and guidance from the Copyright Office, there is significant uncertainty in determining copyright eligibility for AI assisted content, largely because of the specific and nuanced criteria of human input contained within the recent report. Moreover, the exact threshold at which AI-assisted works’ human intervention, such as expressive inputs or detailed creative directions, become sufficient to merit eligibility for copyright protection is unclear. There is yet to be a standard test to know whether copyright protections are warranted; instead, these determinations are made on a case-by-case basis.
Jacob Chang is an Associate Editor on the Michigan Technology Law Review.