Supreme Court Cements Human Authorship Requirement, Exposing Businesses to AI Copyright Liability

The legal foundation of the generative artificial intelligence economy is rapidly solidifying in mid-2026, and the resulting framework places significant new burdens on businesses and creators. On March 2, 2026, the United States Supreme Court quietly denied certiorari in Thaler v. Perlmutter, ending a highly watched, years-long legal battle over whether an artificial intelligence system can hold a copyright. By refusing to hear the appeal, the Court left intact a definitive ruling from the U.S. Court of Appeals for the District of Columbia Circuit, which established that human authorship is a "bedrock requirement" of American copyright law. For companies integrating generative models into their daily workflows, the decision establishes a stark reality: works created exclusively by autonomous software are fundamentally incapable of receiving intellectual property protection.[2][3]

The litigation was intentionally designed as a boundary-testing legal vehicle. Dr. Stephen Thaler, a computer scientist, developed a generative AI system he named the "Creativity Machine." In 2018, the software autonomously generated a piece of visual art titled "A Recent Entrance to Paradise." Thaler submitted a copyright registration application to the U.S. Copyright Office, explicitly listing the Creativity Machine as the sole author and himself as the owner of the machine. He openly conceded in his application that the submission "lacked traditional human authorship," arguing instead that property rights should flow from the machine to its owner under a modernized interpretation of the law.[2][5]

The U.S. Copyright Office firmly rejected Thaler's application in 2019, sparking the federal lawsuit. In its defense, the Office relied heavily on foundational jurisprudence, specifically citing the 1884 Supreme Court decision in Burrow-Giles Lithographic Co. v. Sarony. In that 19th-century case, which determined that photographs could be copyrighted despite being produced by a machine (a camera), the Court ruled that the protection stemmed from the human photographer's intellectual labor—posing the subject, selecting the lighting, and framing the shot. The Copyright Office argued that, unlike a camera wielded by a human, Thaler's Creativity Machine operated autonomously, severing the necessary link between human intellect and the final creative expression.[5]

When the case reached the D.C. Circuit in 2025, the appellate panel conducted a rigorous statutory analysis of the Copyright Act of 1976. Circuit Judge Patricia A. Millett, writing for the court, noted that while the Act does not explicitly define the word "author," its structural provisions inherently assume a human being. For example, the law measures the duration of a copyright by the author's lifespan, requires joint authors to possess mutual intent, and mandates signatures for registration—capacities that only natural persons possess. The court concluded that Congress clearly intended to limit copyright to human creators, effectively closing the door on machine authorship under current statutes.[2][5]

Thaler's defeat at the Supreme Court mirrors a parallel failure in the realm of patent law. Thaler had previously filed patent applications listing another of his AI systems, DABUS, as the sole inventor of a new beverage container and a flashing light beacon. The U.S. Patent and Trademark Office (USPTO) refused the applications, a decision ultimately affirmed by the Federal Circuit in Thaler v. Vidal. The Supreme Court similarly declined to hear that case. To reinforce this stance, the USPTO issued updated guidance in late 2025 explicitly stating that only natural persons can be named as inventors, even in cases where artificial intelligence plays a highly significant role in the inventive process.[4]

While the Supreme Court's denial provides definitive clarity on purely autonomous AI outputs, it leaves a massive legal gray area regarding AI-assisted human creativity. The D.C. Circuit explicitly noted that its ruling does not prohibit copyrighting works made "with the assistance of artificial intelligence." The rule merely requires that the author be the human who "created, operated, or used" the tool. However, because Thaler intentionally disclaimed any human creative input to force a ruling on machine authorship, the courts were not required to establish a standard for exactly how much human involvement is necessary to cross the threshold into copyrightability.[2][5]

This unresolved boundary is currently the source of intense friction between creators and the U.S. Copyright Office. Under current Office guidelines, applicants must explicitly disclose the inclusion of AI-generated content in their works and disclaim protection for those specific components. A human author can copyright the specific arrangement or heavy modification of AI outputs, but the underlying raw generation remains unprotected. Intellectual property lawyers are now advising businesses that they must meticulously document their creative processes—saving prompt histories, iterative edits, and human-directed structural choices—to prove sufficient human authorship in the event of a legal challenge.[1][3][7]

The inability to copyright raw AI outputs creates a severe vulnerability for the corporate sector. If a marketing agency uses an AI tool to generate a comprehensive advertising campaign, or a software firm uses an AI coding assistant to write a proprietary algorithm without meaningful human modification, that content enters the public domain immediately upon creation. According to legal analysts, this means that competitors can legally copy, reuse, or reverse-engineer those assets without facing any copyright infringement penalties. The traditional intellectual property moats that businesses rely on to protect their investments are effectively bypassed.[1][6]

Beyond the loss of protection, businesses utilizing generative tools are now exposed to direct liability risks. Intellectual property attorneys warn that if an AI system generates output that closely mirrors copyrighted training data, the company deploying the tool—not just the developer of the underlying model—is legally responsible for the infringement. The user is considered the responsible party for publishing or commercializing the infringing material. This creates a perilous environment for marketing teams, developers, and content creators who may unknowingly generate and publish protected expressions.[1]

This liability risk is compounding rapidly as the legal battles over the data used to train these models escalate into existential threats for AI developers. In May 2026, a powerful coalition of major publishers—including Hachette, MacMillan, McGraw-Hill, and Elsevier—alongside prominent authors like Scott Turow, filed a landmark class-action lawsuit against Meta. The plaintiffs allege that Meta unlawfully ingested millions of copyrighted books and academic materials to train its Llama series of large language models, specifically pointing to the use of controversial, pirated data repositories.[1][4]

The publishers' lawsuit represents a significant escalation from earlier, smaller-scale creator disputes, bringing the massive legal and financial resources of legacy publishing to bear against the frontier AI industry. The plaintiffs argue that the models serve as commercial substitutes for the original works, capable of summarizing, mimicking, or reproducing the authors' expressions. By ingesting this data without permission or compensation, the publishers claim, AI developers are executing the largest mass copyright infringement in history, directly undermining the economic foundation of the publishing industry.[1][6]

Meta and other leading AI developers are mounting a vigorous defense rooted in the "fair use" doctrine of U.S. copyright law. They argue that training a large language model is a highly transformative process that extracts statistical patterns, syntax rules, and uncopyrightable facts, rather than copying the expressive content of the original works for direct reproduction. The technology industry contends that restricting access to training data would effectively cripple the development of advanced artificial intelligence in the United States, ceding technological leadership to nations with more permissive data-mining laws.[1][6]

The uncertainty surrounding the fair use defense remains the most significant unresolved question in global AI policy. While technology companies frequently point to the early 2010s Authors Guild v. Google case—which permitted Google to scan books to create a searchable database—publishers argue that generative AI is fundamentally different because it generates new, competing market substitutes. Until the Supreme Court or Congress explicitly addresses whether ingesting copyrighted data for model training constitutes fair use, the entire generative AI ecosystem operates under a cloud of profound legal peril.[1][6]

In the interim, the combination of the Thaler precedent and the looming training-data litigation is forcing a rapid shift in corporate behavior. Enterprise software vendors are increasingly offering complex indemnification clauses to protect their enterprise clients from third-party copyright claims, though these often come with strict usage conditions. Simultaneously, corporate legal departments are drafting stringent internal policies governing how employees can use generative tools, mandating human-in-the-loop review processes. The era of unregulated, experimental AI deployment is rapidly closing, replaced by a compliance-driven environment where the provenance of every digital asset must be tracked, verified, and legally defended.[1][3][6]