Appeals Court Rejects Fair Use Defense in Landmark AI Training Copyright Case

First Appeals Court Ruling on AI Training and Fair Use

The Philadelphia-based 3rd US Circuit Court of Appeals delivered a landmark decision on September 30, ruling that defunct legal research company Ross Intelligence did not make fair use of Thomson Reuters' copyrighted material when building its AI-based legal search engine1

1

. This marks the first time a US appeals court has weighed in on fair use in the AI training context, a pivotal defense that tech companies are relying on across dozens of high-stakes copyright infringement lawsuits1

1

.

US Circuit Judge Tamika Montgomery-Reeves wrote for the unanimous three-judge panel that despite Ross' framing of the case as concerning "the future of AI legal technology," the reality was simpler. "But appearances can be deceiving. In truth, this is no more than an ordinary copyright case," she stated2

2

. The court's task was to determine whether Thomson Reuters' materials possessed sufficient originality for copyright protection and whether Ross' use qualified as transformative use under fair use doctrine1

1

.

Thomson Reuters vs Ross Intelligence: The Core Dispute

Thomson Reuters' 2020 lawsuit accused Ross Intelligence of copying thousands of Westlaw headnotes—brief editorial summaries that identify important legal issues in court opinions—to train its competing AI-powered legal research platform3

3

. These headnotes are not part of court decisions themselves but represent independent editorial work created by Westlaw's attorney editors to help lawyers quickly locate relevant portions of judicial opinions2

2

.

Ross developed its system by working with third-party LegalEase Solutions to create training materials connecting legal questions with passages from court opinions, incorporating the Westlaw headnotes in the process2

2

. The company shut down its platform in 2021, citing the litigation costs from Thomson Reuters' lawsuit3

3

.

Why the Court Rejected the Fair Use Defense

The 3rd US Circuit Court of Appeals found that Westlaw's headnotes met the low originality threshold required for copyright protection. Editors had to exercise judgment in deciding which legal issue each headnote would cover, what facts and reasoning to include, and how to describe points accurately and concisely2

2

. "Each headnote had to convey enough factual background and reasoning to stand on its own without reference to the opinion or any other headnote, while accurately reflecting the corresponding opinion passage," Montgomery-Reeves explained2

2

.

Crucially, the court determined that Ross' use was "minimally transformative at best." While Ross took an intermediate step of using the headnotes to train an AI program rather than displaying them directly to users, both Thomson Reuters and Ross ultimately used the headnotes for the same purpose: creating and optimizing a legal research platform5

5

. "Ross took the headnotes to make it easier to develop a competing legal research tool," a Delaware federal court judge had stated in the original ruling3

3

.

Market Harm and Intellectual Property Protection

The appeals court focused heavily on market harm, often the most decisive factor in fair use analysis. The 32-page opinion ruled that Ross' use caused harm to Thomson Reuters by undercutting Westlaw's ability to market its service through a directly competing product5

5

. "By copying the headnotes and using them to build its own competing legal research platform, Ross appropriated the headnotes' value for finding and analyzing judicial opinions and diminished their value as a draw for users to Thomson Reuters' legal research platform," the court stated5

5

.

The ruling also acknowledged potential interference with a future market for licensing headnotes as AI training data2

2

. A Thomson Reuters spokesperson emphasized that "strong copyright protections are essential to promoting innovation and safeguarding the intellectual property that powers our Fiduciary-Grade AI solutions"1

1

.

Distinguishing This Case from Generative AI Disputes

The 3rd Circuit carefully distinguished Thomson Reuters' case from other AI training lawsuits involving generative AI—artificial intelligence that creates new content. Ross' search engine was not generative; it merely returned passages taken directly from judicial opinions in response to user queries5

5

. The appeals court noted in a footnote that concerns raised by the US Department of Justice in a copyright lawsuit against OpenAI "do not apply here"1

1

.

This distinction has sparked debate about the ruling's broader implications. Tech industry group Chamber of Progress senior director of AI Adam Eisgrau suggested the ruling "implicitly confirms that highly transformative gen AI training to produce a hugely multi-purpose model with substantial public benefit is likely fair use"4

4

. However, copyright expert Aaron Moss from Copyright Lately argued the case isn't the "best vehicle for a sweeping ruling on generative AI" given that Ross no longer exists, damages remain unresolved, and its AI wasn't generative4

4

.

What This Means for the AI Industry

The precedent established by this US appeals court ruling arrives as authors, news outlets, music labels, and other rights holders have filed dozens of lawsuits against tech companies over unauthorized use of copyrighted works to train AI systems3

3

. OpenAI CEO Sam Altman has previously maintained that the AI race would be "over" if the fair use provision were overturned4

4

.

The ruling puts added weight on the source and function of copyrighted material for AI training. While AI developers can often obtain public documents like court decisions, this decision clarifies they don't automatically gain rights to use a publisher's summaries, labels, rankings, or other editorial work built around those documents2

2

. The case was based on specific circumstances involving a direct competitor, copyrighted editorial material, and a legal research product performing essentially the same function as the original2

2

.

Looking ahead, cleaner cases for deciding the future of generative AI training are already in motion. A hearing for Concord Music Group's lawsuit against Anthropic—accusing the company of unlawfully using thousands of copyrighted song lyrics to train its Claude AI model—is scheduled for later this month4

4

. The vestiges of Ross have vowed to take their case to the Supreme Court, though whether another decision favoring Thomson Reuters could settle the broader legal battle remains uncertain4

4

.