According to IAM Media, US courts continue to issue inconsistent rulings on whether ingesting copyrighted works to train AI models qualifies as fair use, with outcomes turning heavily on how transformative a model's output is and whether training copies were lawfully acquired. For data licensing desks and AI developers, that means no single precedent yet controls, and litigation risk remains jurisdiction- and fact-specific rather than settled law.
Companies building or licensing training corpora should expect continued case-by-case scrutiny rather than a clean doctrinal rule anytime soon.
How US courts are addressing fair use questions in AI training and copyright disputes
