Lead Article - Generative AI Update: U.S. Courts Address Fair Use Doctrine, Generative AI Authorship, and Patentability
Sign Up for PublicationsU.S. courts are addressing a multitude of issues raised by generative AI. Many cases, like the recently decided Thomson Reuters Enterprise Centre GMBH v. Ross Intelligence Inc., address AI inputs, or, copyrighted materials that are input into the AI system in order the train the AI. Other cases concern generative AI outputs, or, inventions, content, and solutions created by AI systems. This article reviews some recent cases that illustrate U.S. law on fair use doctrine, copyright ownership, and patent eligibility.
Fair Use Doctrine - Thomson Reuters Enter. Ctr. GmbH v. Ross Intel., Inc.
On February 11, 2025, the United States District Court for the District of Delaware issued an opinion in Thomson Reuters Enter. Ctr. GmbH v. Ross Intel., Inc., granting Thomson Reuters’ motion for partial summary judgment and dismissing Ross Intel’s fair use defense as a matter of law. 2025 WL 458520 (D. Del. Feb. 11, 2025). The case has attracted the attention of the AI legal community because it is one of the first to address how the fair use doctrine may apply to copyright infringement claims concerning AI.
The issue arose after Defendant Ross Intel bought approximately 25,000 “Bulk Memos” from LegalEase, A legal outsourcing company, to use as training data for a legal AI search engine they were developing to directly compete with Thomson Reuters’s legal-research platform Westlaw. The “Bulk Memos” were created by lawyers in part by using Thomson Reuters’ Westlaw headnotes—brief summaries of cases’ key takeaways and points of law. Plaintiff Thomson Reuters alleged the use of their copyrights as training data for Ross Intel’s AI search engine directly infringed on their registered copyrights.
While the U.S. District Court in Delaware granted partial summary judgment to Thomson Reuters holding that Ross Intel infringed 2,243 of Thomson Reuters’ headnotes, the most closely watched portion of this case centered around whether Ross’s use of the Bulk Memos to its AI search engine qualified as a fair use. The doctrine of fair use provides that use of another’s work is not infringing when it is used for certain purposes. In determining whether a particular use is fair in a particular case, courts are to consider four non-exclusive factors: “(1) the use’s purpose and character, including whether it is commercial or nonprofit; (2) the copyrighted work’s nature; (3) how much of the work was used and how substantial a part it was relative to the copyrighted work’s whole; and (4) how [the defendant’s] use affected the copyrighted work’s value or potential market.” Id. at *7 (citing 17 U.S.C. § 107(1)–(4)). The Court held that Ross’ fair use defense failed as a matter of law because Thomson Reuters prevailed on factors one and four, “the two most important [factors]…on the overall balancing.” Id.
On factor one, the Court easily determined that the use was commercial both because Ross admitted as much, and because Ross stood to profit from its unauthorized use of the Westlaw headnotes. The Court also held Ross' use was not transformative because Ross did not have a "further purpose or different character from" Thomson Reuters's.” Id. (citing Andy Warhol Found. for the Visual Arts, Inc. v. Goldsmith, 598 U.S. 508, 529 (2023)). The Court further ruled that because the technology at issue was not generative AI (AI that writes new content itself), and instead was designed to produce relevant judicial opinions that have already been written, further analysis would still be required because the copying in question was intermediate—occurring when “Ross turned the headnotes into numerical data about the relationships among legal words to feed into its AI.” Ross, 2025 WL 458520, at *7. The Court rejected Ross's reliance on fair use decisions involving intermediate copying of computer programs by distinguishing the present case from prior cases, such as Google LLC v. Oracle Am., Inc., on two key grounds: (1) those cases involved copying computer code, not written text, primarily for functional purposes; and (2) the intermediate copying in those cases was “necessary” for competitors to access an underlying idea and foster innovation. Id. at *7–8.
As to the fourth factor, the likely effect of Ross’s copying on the current and potential derivative markets for the original work (which has been described as “the single most important element of fair use[,]”), the Court found that because Ross designed their AI search to compete directly with Westlaw, holding in favor of Ross Intel’s fair use affirmative defense would have a significant impact on both the legal research market and the emerging market for training legal AIs. Id. at *9 (quoting Harper & Row Pubs., Inc. v. Nation Enters., 471 U.S. 539, 566 (1985)).
The Court could not find a public benefit that would justify an alternative holding. Rather, the Court acknowledged a “public interest in accessing the law,” but noted that “legal opinions are freely available,” and “[t]here is nothing that Thomson Reuters created that Ross could not have created for itself or hired LegalEase to create for it without infringing Thomson Reuters’s copyrights.” Id. at *10. Weighing all four factors, Judge Bibas granted summary judgment for Thomson Reuters on fair use.
Trial is scheduled to begin the week of May 12, 2025.
For the AI community, the decision indicates that the fair use doctrine may not protect AI developers who use copyrighted content to create competing products, especially when the use is non-transformative and has an effect on the market for the original work. The holding also indicates that moving forward, courts may consider potential future markets for AI training data, not just current market impacts when analyzing a fair use defense.
Generative AI Outputs and Copyright Authorship - Thaler v. Perlmutter
The Copyright Act provides copyright protection to “original works of authorship.” 17 U.S.C. § 102(a). Although the statute does not define the term “author” or “authorship,” courts have permitted only human authors to be eligible for a copyright. The use of machines to aid in the creation of is expressly permitted, though the U.S. Copyright Office (“USCO”) has typically required the “work” to be conceived and executed by a human, with the machine serving as an “assisting instrument” to create the work before it may be registered. U.S. Copyright Office, Sixty-Eighth Annual Report of the Register of Copyrights for the Fiscal Year Ending June 30, 1965, at 5 (1966), https://www.copyright.gov/reports/annual/archive/ar-1965.pdf. The use of Generative AI tools in creating new works poses a novel question of the degree of human intervention needed for the work to be considered authored by humans and not AI.
These questions are currently working their way through U.S. courts. On September 19, 2024, the U.S. Court of Appeals for the District of Columbia heard arguments in the matter of Thaler v. Perlmutter that raised the question of whether an image generated by AI could be registered for a copyright. The appellant, Dr. Stephan Thaler, attempted to obtain a copyright registration for an image that he claimed was autonomously generated by his AI system named “Creativity Machine.” The USCO rejected the application on the basis that the work lacked human authorship. In his First Request for Reconsideration, Thaler confirmed that the work was “autonomously generated” and “lacked traditional human authorship.” First Request for Reconsideration at 2, ECF No. 13-5. In other words, he did not try to argue that humans had a role in creating the work; instead, he contested the longstanding requirement by the USCO that works be registered by a human author arguing that AI can be acknowledged as an author and any copyright ownership in the work should vest with the AI’s owner.
But the Courts and the USCO must also address the harder question of what happens when a human author purports to have had a creative control over the work that incorporates AI output. For example, in Allen v. Perlmutter, a case in the U.S. District Court for the District of Colorado, an artist Jason Allen applied for a copyright registration for an AI-assisted artwork he claims he created using the AI program Midjourney. Unlike Thaler (in which the applicant said AI authored the work), Allen argues that he is the author because he prompted the AI tool to develop images using extremely detailed descriptions and went through several iterations before he considered the work finished. See Complaint ¶ 17, Allen v. Perlmutter, Case No. 1:24-cv-02665 (D. Colo. filed [date] 2024). Allen claims that he is the author and that the AI merely assists in the execution of his creative vision.
The USCO would disagree with Allen. On January 30, 2024, the USCO published Part 2 of the Copyright and Artificial Intelligence report addressing copyrightability in the works made using of generative AI. In the report, the USCO reaffirmed its position from an earlier guidance last year that that “purely AI-generated” works or works lacking “sufficient” human control would continue to be ineligible for copyright. Those works that contain insufficient human control includes the works that are created using “prompts alone.”
Generative AI Outputs and Patent Eligibility – Thaler v. Vidal
In the U.S. and most other countries, new inventions and innovations developed by AI without the assistance of humans are not patentable. Courts in the U.S., the United Kingdom, the EU, Australia, Japan, Korea and New Zealand have explicitly ruled AI cannot be considered an inventor for the purposes of a patent. In 2022, the U.S. Court of Appeals for the Federal Circuit held in Thaler v. Vidal (a parallel case to Thaler v. Perlmutter) that AI could not be listed as the inventor on a patent application. In a scenario similar to Thaler v. Perlmutter, Thaler submitted a patent application alleging an AI system he owns named Device for the Autonomous Bootstrapping of Unified Sentience (aka “DABUS”) “invented” a flickering light that mimics neural activity as well as a drink container designed for easier gripping by robots. The United States Patent and Trademark Office (“USPTO”) denied the application on the basis that the Patent Act “limit[s] inventorship to natural persons.” Thaler v. Vidal made its way to the U.S. Court of Appeals for the Federal Circuit where the parties presented oral argument on June 6, 2022.
Thaler argued that the Patent Act is indifferent as to who or what was the inventor, because its goal is to promote innovation and cited to language in the act which states “[p]atentability shall not be negated by the manner in which the invention was made.” 35 U.S.C. § 103 (2022). Thaler also pointed to industries such as medical and pharma which are increasingly leaning on AI for research and design. To discount their investment by denying patents for their new AI-developed inventions discourages the use of new research tools. Ultimately, the Court of Appeals sided with the USPTO adopting a more textualist approach to interpretation of the Patent Act which expressly provides that inventors be “individuals.” Id. § 100(f). The Court cited a long history of precedent holding that the term “individual” refers to human beings unless there is evidence Congress intended an alternative meeting. Thaler v. Vidal, 43 F.4th 1207, 1211 (Fed. Cir. 2022).
This case illustrates the U.S. courts’ current stance against AI as an independent inventor for a patent. In early 2024, however, the USPTO published inventor guidance on patent eligibility for AI-assisted inventions stating that “the use of an AI system by a natural person(s) does not preclude a natural person(s) from qualifying as an inventor … if the natural person(s) significantly contributed to the invention as the inventor or joint inventors.” Inventorship Guidance for AI-Assisted Inventions, 90 Fed. Reg. 10.043 (Feb. 13, 2024). The USPTO further explained that the significance of the human contribution would be measured by the Pannu factors, which require inventors to “(1) contribute in some significant manner to the conception or reduction to practice of the invention, (2) make a contribution to the claimed invention that is not insignificant in quality, when that contribution is measured against the dimension of the full invention, and (3) do more than merely explain to the real inventors well-known concepts and/or the current state of the art.” Pannu v. Iolab Corp., 155 F.3d 1344, 1351 (Fed. Cir. 1998). As with copyright authorship, as inventors increasingly use AI tools to innovate, the patentability of their inventions will depend on how inventors integrate their own creativity and ideas into prompts for AI as well as the AI’s generated output.
Conclusion
The intersection of generative AI and the law is becoming increasingly complex, with courts addressing pivotal questions around authorship, patentability, and liability. The Thaler cases illustrate the ongoing struggle to define the role of AI within frameworks traditionally built for human actors, particularly in copyright and patent law. Although U.S. courts have largely reaffirmed the necessity of human involvement for intellectual property protections, these rulings underscore the challenges in adapting existing statutes to account for the use of AI tools in human creativity and innovation.
As AI systems become more sophisticated, their transformative potential will continue to push the boundaries of existing legal frameworks. Courts and regulatory bodies face the critical task of ensuring that innovation is balanced with ethical considerations and fairness. The outcomes of these cases, along with emerging legislative guidance, will continue to shape the future landscape of generative AI and its role in society.