According to AIbase, after tech giants such as Meta, Apple, and Anthropic, software giant Adobe is now also caught in an artificial intelligence copyright dispute. Recently, Oregon author Elizabeth Lyon filed a class-action lawsuit, accusing Adobe of using an illegal dataset containing a large number of pirated books to pre-train its small language model series SlimLM.

Adobe Firefly, firefly, generative AI, artificial intelligence, image generation

As an important part of Adobe's AI strategy, SlimLM is designed to optimize document assistance tasks on mobile devices, but the plaintiffs pointed out that the open-source dataset SlimPajama-627B, on which the model relies, is actually a derivative copy of RedPajama, which has been widely criticized for including the "Books3" database with 191,000 infringing books.

Lyon emphasized in the lawsuit that her multiple non-fiction writing guides were included in the training without authorization, without attribution, and without compensation, directly violating copyright law. Although Adobe has previously promoted its AI tools like Firefly as being built on legal, protected content, this accusation against SlimLM has revealed potential compliance risks in its technology infrastructure.

Currently, the tech industry is facing a wave of legal challenges. Anthropic previously paid a $1.5 billion fine for similar issues. With stricter regulations, Adobe's case could become another key turning point in evaluating the legality of "inspiration programming" and AI content generation.