EFF: Rewriting Copyright Law to Fight Generative AI Would Hurt Everyone

Publishers and creators across the US are suing large AI companies for training generative models on copyrighted works. These cases are easy to read as a David-and-Goliath contest, with authors demanding a share of Big Tech profits. The Electronic Frontier Foundation (EFF) argues that far more is at stake. The outcome will affect billions of people who have no seat in these cases: researchers, artists, activists, journalists, people with disabilities, and anyone who accesses or shares content online.

That is why EFF has filed an amicus brief in In re OpenAI. The brief explains how the court's fair use ruling could reach people worldwide. According to authors Corynne McSherry and Tori Noble, a narrower fair use doctrine would impede medical research and investigative journalism and could even limit access to privacy protection tools.

EFF stresses that creators and workers deserve support as AI reshapes their fields, and that Big Tech must answer for the harms it causes. Giving rightsholders more control over the web, however, would achieve neither goal.

The plaintiffs' arguments and their consequences

Fears that AI models would become "infringement machines," producing near-copies of their training data, have not been borne out. Courts have already found that training is fair use, because it puts the works to a fundamentally different purpose.

The plaintiffs respond that even transformative use is unfair, since models can generate new, non-infringing works that compete with the originals. They also argue that collecting data without permission, sometimes from sites hosting infringing copies, is itself infringement.

EFF points out that journalists and nonprofit researchers gather data through the same crawling and scraping techniques. They use them, for example, to investigate possible anticompetitive practices by tech firms or misinformation and hateful content on X, with far less money than the AI giants. Requiring proof that material was lawfully obtained, the brief notes, would cast legal doubt on the billions of copies we all make every day by opening emails, visiting websites, streaming music and watching videos.

Copyright, moreover, exists to encourage new creation, not to shield those who published first. The plaintiffs' reasoning could block books inspired by a favorite genre, or academic literature reviews. EFF also highlights the value of general-purpose models: Be My AI, built on GPT, describes images for people who are blind or have low vision, and generative tools are helping researchers develop new treatments.

Licensing would entrench the giants

In EFF's view, mandatory licensing would hand the largest companies a huge advantage over small competitors and hinder academic research. As Professor Lawrence Lessig has observed, if all training requires a license, only a handful of tech giants could afford it. The cause is not greedy authors but a copyright regime that is "the least efficient property system known to man."

Creators themselves are unlikely to benefit. Cory Doctorow likens giving them more copyright to helping a bullied kid by handing over more lunch money for the bully to take. Labels and publishers already offer artists poor deals.

EFF concludes that the rules for AI should serve the public, not the few companies able to litigate for decades or settle for billions.

Source: EFF Deeplinks