Court filings show Meta staffers discussed using copyrighted content for AI training
Meta staff discussed using copyrighted works for AI training, shown in court filings for Kadrey v. Meta.
Court documents unsealed in Kadrey v. Meta reveal internal discussions among Meta employees about using copyrighted works to train AI models. The lawsuit, involving authors like Sarah Silverman and Ta-Nehisi Coates, disputes Meta's 'fair use' argument for using protected intellectual property.
The discussions include ideas such as purchasing e-books at retail price to avoid lengthy licensing agreements, and contemplating using unauthorized platforms, like Libgen, for training data. Despite facing legal risks, Meta's team has been less conservative in seeking approvals now, with more resources and legal support at their disposal.
Meta is said to have potentially used data from Reddit and examined incorporating other platforms like Quora. The company has even considered overriding previous decisions on using specific datasets, emphasizing a need for sufficient data to compete in AI. Meta's defense in the legal battle has been bolstered by the inclusion of Supreme Court litigators.