How Generative AI Litigation Is Changing the Meaning of “Fair Use”

Courts across the United States are grappling with whether long-standing copyright rules fit the new reality of machine learning. The outcomes of these cases will shape not only how AI systems are built but also how authors and creators are compensated. The most pressing question centers on fair use: does training algorithms on copyrighted material qualify as a protected use, or does it amount to unlawful copying?

The stakes are massive. If courts reject broad interpretations of fair use, technology companies may need to completely rethink how they gather training data, while content owners could gain new leverage to demand licensing fees.

The Thomson Reuters Case: A Turning Point

In February 2025, a federal judge in Delaware issued a closely watched opinion in Thomson Reuters Enterprise Centre GmbH v. ROSS Intelligence Inc. The case involved an AI startup that used Westlaw headnotes to build a legal research tool. The court concluded that ROSS infringed on Thomson Reuters’ copyrights, finding that its reliance on fair use was legally insufficient.

The ruling highlighted that copying large volumes of headnotes—editorial summaries written by Westlaw staff—was not justified under the doctrine. The judge emphasized the competitive relationship between ROSS’s product and Westlaw’s own services. Market harm, traditionally the fourth factor of the fair use test, was deemed the most significant issue, and the decision strongly favored Thomson Reuters.

This analysis disrupted the widespread belief in the tech sector that “publicly accessible” data can automatically be used for training AI.

Rethinking Transformative Use

For years, companies have argued that training a model transforms copyrighted content into something fundamentally different. Courts, however, are questioning this logic. Judges are now asking whether the training itself serves a truly distinct purpose from the original material, rather than focusing only on the outputs an AI system generates.

In the Thomson Reuters case, the court analogized the creation of headnotes to artistic judgment, rejecting the idea that their factual connection to judicial opinions stripped them of protection. This approach broadens what courts may view as creative expression, increasing the scope of materials shielded by copyright.

Applying Four Factors to Fair Use Rulings

When analyzing fair use, courts traditionally weigh four factors. In the AI context, each one has taken on new importance:

• Purpose and character of the use: AI systems are generally developed for commercial gain, a fact that weighs against fair use—particularly when the AI product competes with the original work.

• Nature of the work: Even materials based on factual information may receive strong protection if they involve editorial choices or creative framing.

• Amount used: Training typically involves wholesale copying of entire works, not excerpts. Courts view this as excessive.

• Market impact: If an AI tool substitutes for the original product, market harm is presumed. This has become the deciding factor in many rulings.

Industry Reactions

In light of these rulings, some AI developers are abandoning reliance on fair use and instead negotiating licenses with content owners. Media groups, publishers, and news organizations are beginning to license libraries of material to AI companies, providing a clearer legal basis for training data. While these agreements reduce litigation risk, they also add significant costs, which could reshape the economics of the AI industry.

AI firms are also advancing fresh arguments beyond traditional fair use. Some contend that training merely extracts statistical patterns and does not reproduce works in a meaningful sense. Others emphasize public benefits of AI innovation, likening training to research rather than commercial exploitation.

Regulators, however, have hinted otherwise. The U.S. Copyright Office has suggested that ingesting copyrighted material to train AI could itself be an infringing act. If model outputs bear resemblance to training data, that risk grows stronger.

Different Applications, Different Risks

Not all AI tools face the same level of legal exposure. Text generation engines raise different issues than image synthesis or code completion systems. Legal research platforms like ROSS face intense scrutiny because they directly challenge the markets of established copyright holders. Courts are becoming increasingly nuanced in tailoring their rulings to the technology at hand.

Contrasting Decisions: Anthropic Lawsuit

In a separate case, the AI company Anthropic successfully settled copyright claims brought by authors. That ruling offered a different take on fair use, showing how fact-specific these cases can be. Together, the contrasting outcomes suggest that appellate courts—or perhaps even the U.S. Supreme Court—may eventually need to provide nationwide guidance.

Decisions on fair use in multiple other AI cases are forthcoming next year. Given the liability risk and unpredictability, recent rulings could provide an impetus to settle, experts say.

Historical Echoes: Copyright Battles in Music and Publishing

The current AI disputes echo earlier fights in other industries. Musicians, publishers, and software developers have long battled over where to draw the line between inspiration and infringement. The AI cases may prove to be the modern equivalent of sampling disputes in music or digital copying controversies from the early internet era, setting precedents that define creative rights for decades.

What Comes Next

The next few years will be pivotal. Developers of generative AI must carefully evaluate training datasets and may need to budget for licensing fees to avoid liability. At the same time, authors, musicians, and publishers are gaining stronger negotiating positions to secure payment for the use of their works.

Ultimately, these disputes illustrate the enduring tension between innovation and intellectual property protection. Without legislative guidance, courts are left to reshape fair use one ruling at a time. The trajectory suggests that AI companies must prepare for more restrictive interpretations of fair use, while rights holders should expect expanded opportunities to enforce and monetize their copyrights.

Why Hire The Lyon Firm

AI copyright lawsuits are not only legally complex but also evolving faster than most areas of intellectual property law. For creators, musicians, publishers, and businesses, navigating this uncharted landscape requires experienced counsel. The Lyon Firm provides:

• Deep Knowledge of Emerging Tech Law: We stay ahead of the curve on court rulings and regulatory developments in AI, copyright, and fair use disputes.

• Proven Litigation Experience: Our team has a track record of representing individuals and companies in high-stakes intellectual property cases.

• Personalized Strategy: Each AI dispute is unique, and we tailor our legal approach to the facts of your case, whether you’re protecting creative works or defending against claims.

• Commitment to Justice: We believe creators deserve to be compensated for their work, and businesses deserve clarity in how they can innovate without undue risk.