Adobe Faces Class-Action Lawsuit Over Alleged Use of Authors’ Works in AI Training
Adobe, the tech giant known for its creative software, is facing a proposed class-action lawsuit alleging the misuse of authors’ works to train its AI systems. The suit, filed on behalf of Oregon author Elizabeth Lyon, claims Adobe incorporated pirated books into the training dataset for its SlimLM program, an AI-powered language model designed to assist with document tasks on mobile devices.
According to the lawsuit, Adobe’s SlimLM was trained using SlimPajama-627B, a dataset derived from Cerebras’ open-source collection. Lyon, who has authored several non-fiction writing guides, alleges that her works were included in this dataset without consent. The complaint asserts that SlimPajama is a manipulated version of RedPajama, which itself contains Books3, a massive compilation of over 190,000 copyrighted books used in AI training.
The legal filing highlights the broader challenge facing AI companies: the use of copyrighted material without authorization. Lyon’s attorneys argue that Adobe’s SlimLM training process involved reproducing her copyrighted works, potentially infringing on her rights and those of other authors included in the dataset. This case echoes previous litigation against major tech firms over similar practices.
In recent months, AI training lawsuits have become increasingly common. Apple and Salesforce both faced claims involving RedPajama and Books3, while Anthropic recently settled a $1.5 billion lawsuit with authors over alleged unauthorized use of their works in training its AI chatbot, Claude. These cases underscore growing concerns over copyright, consent, and compensation in the AI industry.
As AI tools continue to evolve and integrate into everyday software, Adobe’s lawsuit may serve as a bellwether for how courts handle intellectual property rights in the era of generative AI. Authors and creators are watching closely, and the outcome could shape the future of AI training practices across the tech industry.
source: Techcrunch