Training AI Using 'Pirated' Content Can Be Fair Use, Law Professors Argue

a year ago

Tech companies, including Meta, are being sued for using copyrighted content as training data for LLMs without permission.
Meta faces a class action lawsuit from authors for allegedly downloading pirated books via BitTorrent and using them as training material.
Meta argues fair use, while authors claim using pirated content disqualifies fair use.
IP law professors filed an amicus brief supporting Meta's fair use defense, emphasizing transformative use in creating AI models.
The brief cites cases like Perfect 10 v. Amazon to argue that transformative, non-expressive use favors fair use.
The professors distinguish Meta's case from direct consumptive use, focusing on internal copying for AI development.
The brief suggests copyright shouldn't hinder technological innovation, comparing AI to past tech advancements like VCRs.
The case highlights global differences, with countries like Japan allowing copyrighted material for AI training under exceptions.
The amicus brief was granted by the court and is publicly available.

Hasty Briefsbeta