Author Elizabeth Lyon has filed a lawsuit that puts Adobe squarely in the same crosshairs as OpenAI and Anthropic.
Filed in California federal court, this class-action lawsuit could fundamentally rewrite how tech giants source training data for AI development.
Lyon, who writes instructional guides for authors, claims Adobe used unauthorized copies of countless books to train its SlimLM language models without permission or payment, according to the Reuters report.
The proposed class action represents potentially thousands of copyright holders whose works were allegedly misused through the Books3 dataset.
A land of lawsuits
Adobe’s SlimLM models were trained using the massive SlimPajama-627B training dataset, which allegedly contains material from the Books3 collection of 191,000 titles. This marks Adobe’s first major copyright lawsuit directly targeting its AI practices, while companies like OpenAI, Meta, Google, and Anthropic have been fighting these battles for months.
The case reveals something potentially devastating for AI developers: using “open-source” datasets can inadvertently introduce copyrighted material, creating massive liability exposure down the line. Courts will now determine how responsibility flows through the data supply chain—specifically, whether companies training on third-party datasets can be held accountable for unauthorized works included by original compilers. This liability question could fundamentally alter how AI companies approach data sourcing.
Billion-dollar precedent
Recent developments make Lyon’s lawsuit especially dangerous for Adobe. Three months ago, Anthropic agreed to pay $1.5 billion to settle similar copyright claims—approximately $3,000 per affected book. That groundbreaking settlement represents the largest publicly reported copyright recovery in history and establishes a precedent requiring AI companies to compensate copyright owners.
The Anthropic case involved a federal judge ruling that while training AI on copyrighted books might qualify as fair use, maintaining pirated copies crossed legal boundaries. With statutory damages for willful copyright infringement reaching up to $150,000 per work, the potential financial exposure is staggering. Consider this: if Adobe faces similar liability across thousands of works, the damages could reach hundreds of millions or even billions of dollars.
Surprise!
This legal earthquake is already reshaping industry practices. This lawsuit creates growing demand for vendors offering fully licensed datasets or tools to track exactly where training data comes from, with companies potentially benefiting from heightened copyright risk awareness. Enterprises are increasingly prioritizing datasets with clear licensing, while publishers continue negotiating rights agreements with AI developers.
Surprisingly, Adobe’s stock actually rose 1.9% despite the lawsuit announcement, suggesting investor confidence in the company’s long-term prospects. However, this case may force the entire AI industry to implement stricter controls over data origins and purge unauthorized works from training libraries.
The UK government just suffered a crushing defeat in its controversial plans to let AI companies freely train on copyrighted work.