Judge Reverses AI Copyright Ruling, Rejects Fair Use Defense in Thomson Reuters v. ROSS

On February 13, 2025, a significant legal development occurred in the ongoing battle over AI and copyright law. Third Circuit Judge Stephanos Bibas, sitting by designation in the District of Delaware, reversed his earlier ruling and granted summary judgment to Thomson Reuters in its lawsuit against ROSS Intelligence. The case, Thomson Reuters v. ROSS, 1:20-cv-00613-SB, is notable for being the first district court decision to address fair use and copyright infringement related to training AI models.

Why it Matters

This ruling has far-reaching implications for the AI industry, particularly in how companies can use copyrighted material to train their algorithms. The decision hinges on the specific context of ROSS Intelligence's actions but serves as a cautionary tale for other developers. Judge Bibas's reversal underscores the complexity and evolving nature of copyright law in the digital age.

Background

In 2020, Thomson Reuters (Reuters) filed a lawsuit against ROSS Intelligence (ROSS), alleging that ROSS had knowingly, intentionally, and willfully infringed on its copyrights by using critical features of Westlaw to develop a competing legal research tool without permission or compensation. The case proceeded through the judicial process, with initial summary judgment motions being denied in 2023.

Just before trial was set to begin in August 2024, Judge Bibas made an unexpected move. He invited both parties to renew their summary judgment motions, a decision that surprised legal observers. Reuters and ROSS complied, leading to the recent ruling.

The Ruling

On February 11, 2025, Judge Bibas issued his decision on the renewed motions for summary judgment. In a significant reversal from his earlier stance, he ruled in favor of Thomson Reuters on both copyright infringement and fair use.

Copyrightability of Westlaw Content

Judge Bibas refined his earlier analysis to conclude that Westlaw's headnotes and Key Number System meet the minimal threshold for copyrightability due to the creativity required to distill, synthesize, or explain legal concepts. This finding is crucial as it establishes a precedent for protecting editorial content in digital databases.

Rejection of Fair Use Defense

The court rejected ROSS’s fair use defense based on several key factors:

1. Competitive Intent: The court found that ROSS copied Westlaw's headnotes to develop a competing product, which undermines the transformative nature required for a fair use claim.
2. Nature of the Copied Material: Unlike computer code, the written words in Westlaw’s headnotes are protected by copyright. The court determined that ROSS’s copying was not necessary to achieve its new purpose.
3. Intermediate Copying Defense: The court dismissed ROSS's argument that its copying was an intermediate step in creating a new product. This defense is more applicable to cases involving computer code, not written content.

Key Risks

The ruling poses several risks for AI developers and companies using copyrighted material:

• Legal Uncertainty: While this case does not definitively resolve the fair use question for all AI applications, it introduces significant legal uncertainty. Developers must carefully consider the copyright implications of their training data.
• Increased Litigation: The decision may encourage more copyright holders to pursue legal action against AI companies, leading to a proliferation of lawsuits and potential chilling effects on innovation.
• Regulatory Scrutiny: This ruling could prompt regulatory bodies to take a closer look at AI training practices, potentially leading to new guidelines or regulations.

The Opportunity

Despite the risks, this case also presents opportunities:

• Clearer Guidelines: The decision provides clearer guidance on what constitutes fair use in the context of AI training, helping developers navigate the legal landscape more effectively.
• Innovative Solutions: Companies may be spurred to develop alternative methods for training AI that do not rely on copyrighted material, fostering innovation in data generation and synthesis techniques.

Conclusion

The reversal by Judge Bibas in Thomson Reuters v. ROSS marks a significant moment in the legal debate over AI and copyright. While it does not close the door on fair use defenses for all AI applications, it highlights the need for careful consideration of intellectual property rights. As the landscape continues to evolve, stakeholders must stay informed and adapt their