Adobe is facing a proposed class-action lawsuit filed by Oregon author Elizabeth Lyon, who alleges that the company used pirated books, including her own works, to train its AI model, SlimLM. The lawsuit claims that Adobe’s training dataset, called SlimPajama, is a derivative of the RedPajama dataset, which contains copyrighted material. Lyon argues that her writing was included in this manipulated dataset without her consent. SlimLM is described by Adobe as a small language model designed for mobile document assistance. The legal issues arise from the use of the SlimPajama dataset, which reportedly incorporates the Books3 dataset, a significant collection of 191,000 books that has been central to multiple legal disputes in the tech sector. Similar accusations have been made against other companies, such as Apple and Salesforce, for allegedly using copyrighted material without permission in their AI training processes. The trend of lawsuits against tech companies over AI training data is becoming increasingly common. In September, Anthropic settled a case for $1.5 billion with authors who accused it of using their pirated work to train its chatbot, Claude. These ongoing legal battles highlight the complex intersection of copyright law and AI development, raising questions about the ethics of using existing works in training datasets.