Authors allege that Meta Platforms ignored legal counsel and employed copyrighted books to train its AI, disregarding warnings from its own legal team. The lawsuit, led by comedian Sarah Silverman and Pulitzer Prize winner Michael Chabon, asserts that Meta used their intellectual property without authorization for its AI model, Llama. Recently submitted evidence, including chat logs, implies Meta’s awareness of potential copyright concerns tied to using the dataset to train Llama.
Discord conversations involving Tim Dettmers, a researcher associated with Meta, reveal talks with the company’s legal department about the legality of utilizing book files for training purposes. Although specifics of the legal issues remain unspecified, there are references to concerns about copyrighted books.
Why does this matter?
This legal dispute, part of several faced by technology companies, might compel firms to compensate creators for using their content to train their products, as a result it may impact the AI sector by potentially elevating the expenses of creating data-centric models. Meta’s rollout of Llama 2, with undisclosed training data, exacerbates this legal scrutiny, especially in light of evolving AI regulations in Europe.