Westlaw Wins: Court Rules AI Training Can Infringe Copyright

At the heart of this dispute is Westlaw’s headnotes—editorial summaries of legal principles extracted from case law. Thomson Reuters, which owns Westlaw, claims copyright over these headnotes and argued that Ross Intelligence copied them without permission to train its AI-driven legal research platform.

Ross, in turn, attempted to license the data but was denied, leading it to seek alternative means to develop its AI. The company partnered with LegalEase, a third-party legal research firm, which created "Bulk Memos"—a dataset built using Westlaw headnotes to generate legal questions and answers for AI training. When Thomson Reuters discovered this, it sued for copyright infringement.

The legal question before the court was straightforward but complex in its implications:

Did Ross Intelligence illegally copy protected Westlaw content, or was this a fair use of legal data?

The Court’s Decision: Why Westlaw Prevailed

Initially, in 2023, the court denied summary judgment, allowing the case to move forward. However, upon a closer examination of the facts and legal arguments, the judge reversed course and ruled in favor of Thomson Reuters in February 2025.

Key Findings:

• Ross Intelligence directly copied 2,243 Westlaw headnotes to train its AI, as proven through expert testimony and textual comparison.

• The court rejected Ross’s "fair use" defense, finding that its use was both commercial and non-transformative—meaning Ross wasn’t significantly altering or repurposing the content.

• Ross’s additional defenses failed (e.g., innocent infringement, copyright misuse, merger doctrine, and scenes à faire).

• The ruling leaves unresolved questions about expired copyrights, which will be determined at trial.

Fair Use: Why Ross’s Defense Failed

Ross Intelligence relied on the "fair use" doctrine under 17 U.S.C. § 107, but the court ruled against it based on the four-factor test:

Legal and Business Implications of This Decision

1. AI Training & Copyright Law: The Need for Clearer Guidelines

This case highlights the legal gray area surrounding AI training data. Unlike cases involving traditional software (Google v. Oracle), this ruling makes it clear that using copyrighted legal texts for AI training is not automatically protected under fair use.

What this means for legal tech companies:

• AI firms must be cautious when using proprietary databases like Westlaw or LexisNexis.

• Licensing agreements may become mandatory for AI training datasets.

• The case sets a precedent for future AI copyright battles, especially in legal research

2. Copyright Protection in Legal Research Tools

While judicial opinions are public domain, editorial elements like headnotes, key number systems, and research annotations can be copyrighted. This ruling strengthens Thomson Reuters’ position and could impact:

• Startups developing AI-powered legal research tools—they may need to build their own datasets.

• Law firms using AI tools—the risk of secondary liability for using AI trained on copyrighted material.

3. The Future of AI in the Legal Industry

Ross Intelligence aimed to disrupt the legal research market with AI-driven tools. However, this case suggests that AI companies must either create original datasets or license legal content legally.

• Could this ruling stifle AI innovation? Some argue that limiting AI’s ability to train on existing legal texts could slow legal tech advancements.

• Will we see new legal battles over AI-generated content? As AI systems become more advanced, the debate over ownership, originality, and copyright will intensify.

  
  

Final Thoughts: What Lawyers & Legal Tech Professionals Should Watch For

The Thomson Reuters v. Ross Intelligence decision represents a significant turning point in how copyright law applies to AI-driven legal research tools. This case highlights the increasing tension between technological innovation and intellectual property protection in the legal industry. AI companies cannot assume that using copyrighted materials for training purposes will automatically fall under fair use, especially when the resulting product competes with the original content provider.

The ruling reinforces copyright protections for legal research tools, signaling potential challenges for AI developers who rely on existing legal texts to train their models. Law firms and legal professionals must also be cautious when adopting AI-based research tools, as using systems trained on copyrighted content could create liability risks. As AI continues to shape the legal profession, courts will play an essential role in defining the limits of copyright protection and the scope of fair use in this rapidly evolving area.

What’s Next?

The next phase of this case will determine whether Ross Intelligence appeals the ruling or if the trial proceeds to resolve outstanding issues, such as whether some of the copied headnotes fall into the public domain. Beyond this case, legal tech companies must consider how they will navigate copyright restrictions when developing AI-based legal research tools. One possible outcome is that companies like Westlaw and LexisNexis may begin licensing their data for AI training, creating new revenue models while ensuring legal compliance.

The broader question is whether this ruling will set a precedent for future AI-related copyright disputes, particularly in industries that rely heavily on curated, structured databases. Lawyers, legal tech entrepreneurs, and AI developers should closely monitor how courts continue to address these issues, as the legal framework governing AI and copyright is still in its early stages.