Sarah Silverman sues OpenAI and Meta alleging copyright infringement

Sarah Silverman sues OpenAI and Meta alleging copyright infringement


The comedian Sarah Silverman, along with two other authors, are suing Meta (the maker of ChatGPT) and OpenAI (the maker of ChatGPT), alleging that the AI language models used by both companies were trained using copyrighted material from their books.

Both lawsuits, which seek class action status, were filed Friday in San Francisco Federal Court against OpenAI and Facebook parent Meta. Silverman, author of "The Bedwetter," is joined by Christopher Golden and Richard Kadrey in filing these lawsuits.

In recent months, a new generation of AI tools have gained a lot of attention for their ability generate images and written content in response to prompts from users. These tools are based on large language models that have been trained using vast amounts of online data. This practice has led to some concern that the models could be using copyrighted work without permission.

According to the complaint, 'when ChatGPT prompts are given, ChatGPT generates summary of Plaintiffs copyrighted work--something that is only possible if ChatGPT has been trained using Plaintiffs copyrighted work.' According to the complaint, the authors "did not consent" to ChatGPT using their copyrighted works as training material.

Meta is also accused of using the copyrighted works of authors to train LLaMA (the set large language models that Meta released in February). The lawsuit claims that Meta used copyrighted books, including those written by plaintiffs, to train its language models.

The lawsuit against Meta alleges also that the company obtained the copyrighted materials via an online "shadow library" website, which includes a huge quantity of copyrighted content.

Meta has declined to comment about the lawsuit. OpenAI didn't immediately respond to our request for comment.

Silverman's legal action isn’t the first one to examine how large language models can be trained. In a separate lawsuit, filed last month against OpenAI, the company was accused of misappropriating vast swaths personal data on the internet in order to train its AI tools. (OpenAI has not responded to a request for a comment on the lawsuit.)

OpenAI CEO Sam Altman acknowledged in May that more needs to be done for creators to be satisfied with how AI systems use the works they create.

He said that he was working on a new model where you would be paid if an AI system used your content or style.