Judge sides with Meta in one AI training copyright case, with limits

In short: A US judge gave Meta a win in one copyright lawsuit over using books to train AI, but said the decision is limited and does not legalize all AI training.

What happened

A judge ruled for Meta at the summary judgment stage in a copyright case brought by 13 authors. Summary judgment means the judge decided the issue based on the written record, without a full trial.

The court said Meta’s copying of the books for training its AI models was “fair use” on the facts presented. Fair use is a part of US copyright law that can allow limited copying for certain purposes, like criticism, research, or other uses the law sees as meaningfully different from the original.

But the judge also stressed the ruling is narrow. It applies only to the 13 authors who sued and the specific arguments and evidence in that case. It does not mean every book Meta used was covered, or that all AI training by Meta or other companies is automatically lawful.

Why it matters

This decision matters because it signals that some courts may accept fair use arguments for AI training, at least in certain situations. A big focus is “market harm,” which is basically whether the copying hurts the author’s ability to sell books or licenses, like a knockoff that replaces the real product. Commentators say it can be hard for authors to prove that training data caused lost sales, especially when the AI does not spit out the original text.

The ruling also leaves major questions open, including whether the training data was obtained legally. Using pirated books can raise more legal risk, even if the training process itself is treated as fair use. Future lawsuits with stronger evidence on piracy or lost licensing income could lead to different outcomes.

Source: NYTimes