A federal district court in Delaware has issued the first AI copyright fair use decision on the merits, granting partial summary judgment for copyright owner Thomson Reuters on copyright infringement and rejecting defendant Ross Intelligence’s fair use defensei. The case has important implications for the dozens of other AI copyright cases that are working their way through the U.S. courts.
Background
This closely-watched case, Thomson Reuters v. Ross Intelligence, Inc., has been pending since 2020, well before ChatGPT launched the boom in generative AI chatbots and tools.
The plaintiff Thomson Reuters publishes the Westlaw legal research platform. It owns copyright registrations in case headnotes and in the Key Number System numerical taxonomy for organizing its content.
The defendant Ross offers a legal-research search engine that uses artificial intelligence. To train its AI, Ross obtained training data from a third party. That data, in turn, was based on Westlaw’s headnotes.
Thomson Reuters sued Ross for copyright infringement. In 2023, Judge Stephanos Bibas, a Third Circuit Court of Appeals judge sitting by designation in the District of Delaware, denied motions for summary judgment, finding that disputed issues of fact required trial on infringement and fair use. But with trial scheduled for August 2024, Judge Bibas reconsidered his ruling, continued the trial, and invited the parties to renew their summary judgment motions.
On February 11, 2025, the court issued its opinion:
- Granting partial summary judgment for Thomson Reuters on direct copyright infringement; and
- Denying summary judgment for Ross on fair use.
Infringement ruling
In granting partial summary judgment to Thomson Reuters, the court ruled that 2,243 of the headnotes satisfied the originality requirement and that Thomson Reuters owned valid copyrights in them. The court then found that Ross engaged in actual copying of the headnotes and that Ross’s works were substantially similar.
The court left for trial whether Ross infringed other works, such as the Key Number System, editorial decisions in 500 judicial opinions, and thousands of other headnotes.
The court rejected Ross’s defenses of innocent infringement, copyright misuse, merger, and scenes à faire.
Fair use ruling
Turning to Ross’s fair use defense, Judge Bibas considered each of the four fair use factors. Weighing them together, the court found no fair use.
- Factor 1 – Purpose and Character of the Use – for Thomson Reuters. Ross’s use was commercial, and it was not transformative. Ross used the headnotes “as AI data to create a legal research tool to compete with Westlaw.” The court noted that Ross’s AI “is not generative AI (AI that writes new content itself).” “Rather, when a user enters a legal question, [Ross] spits back relevant judicial opinions that have already been written.”
It did not change the result that Ross’s copying occurred at an intermediate step – specifically, that Ross “turned the headnotes into numerical data about relationships among legal words to feed into its AI.” That is because the cases finding intermediate copying to be fair use were about copying computer code, which is not what Ross did. Such intermediate copying was not necessary for Ross to innovate: “there is no computer code whose underlying ideas can be reached only by copying their expression.” Ross “took the headnotes to make it easier to develop a competing legal research tool.”
- Factor 2 – Nature of the Copyrighted Works – for Ross. The asserted headnotes and Key Number System reflected limited creativity. This weighed in favor or fair use.
- Factor 3 – Amount and Substantiality of the Portion Used – for Ross. Ross’s output to an end user did not include a Thomson Reuters headnote. That weighed in favor of fair use “[b]ecause Ross did not make West headnotes available to the public.”
- Factor 4 – Effect on the Market for or Value of the Copyrighted Works – for Thomson Reuters. Ross “meant to compete with Westlaw by developing a market substitute.” The use harmed both the market for legal-research platforms and the potential derivative market for data to train legal AIs.
Takeaways
Judge Bibas’s decision has important implications for fair use jurisprudence and for the dozens of AI copyright cases around the country.
- The decision raises questions about the strength of fair use defenses in AI cases. That especially is the case where, as here, the material copied is not computer code, and where the copying was done at an intermediate step to train an AI platform—a common feature across AI tools.
- It may be even more challenging to establish fair use where the copyrighted works are highly creative. While Factor 2 rarely drives the analysis, it will weigh in favor of copyright owners, and against fair use, where the asserted works are novels, music, images, and other creative works.
- This case did not decide fair use for generative AI. Importantly – and Judge Bibas took pains to note it – the decision did not involve a generative AI tool. Other courts will have to take up the crucial questions of transformativeness and market effects in the context of generative AI.
Fair use ultimately is a fact-driven inquiry. Each case must be considered on its unique record. While Thomson Reuters is the first case to tackle the merits of the fair use defense for AI, it certainly will not be the last.
i Thomson Reuters Enter. Centre GmbH v. Ross Intelligence, Inc., No. 1:20-cv-613-SB (D. Del. Feb. 11, 2025).
This informational piece, which may be considered advertising under the ethical rules of certain jurisdictions, is provided on the understanding that it does not constitute the rendering of legal advice or other professional advice by Goodwin or its lawyers. Prior results do not guarantee similar outcomes.