AI Rights Project challenges AI fair use reasoning in Mosaic LLM case

The AI Rights Project on August 26 asked a federal judge in California for permission to file an amicus brief in In re Mosaic LLM Litigation, arguing that recent AI copyright rulings rest on an untested assumption about machine training and human reading. The group says the issue is big enough that Congress, not courts, should decide whether fair use extends to machine-scale training.

Why it matters: - Two Northern District of California rulings have said training an AI model on lawfully obtained copyrighted works is noninfringing fair use. - Those decisions allow AI developers to use copyrighted material for training without permission or payment. - The AI Rights Project says that result could turn on a legal assumption no court has actually tested.

What happened: - The AI Rights Project moved on August 26 for leave to file an amicus curiae brief in In re Mosaic LLM Litigation, No. 3:24-cv-01451-CRB, before Judge Charles R. Breyer in the Northern District of California. - The filing is the first under the group’s amicus program. - The program applies the group’s Human Baseline Framework to precedent-setting AI litigation.

The details: - The group’s brief says courts have treated machine training as equivalent to human reading in two California rulings. - In Bartz v. Anthropic PBC, the court described training as the statistical discovery of relationships among word fragments and compared it to “any reader aspiring to be a writer.” - The AI Rights Project argues the courts did not examine whether machine training and human reading are alike in the way that matters for fair use. - The brief says fair use has historically assumed a person exercising creative judgment and relating copied material to a new purpose. - The filing does not argue that fair use fails outright. - Instead, it argues that extending fair use to machine-scale training is a decision for Congress, not one for a court to make by analogy. - Jim W. Ko, founder and executive director of The AI Rights Project, said a premise “nobody has examined” is now doing major work in U.S. copyright law. - The AI Rights Project says it has no stake in which party wins the case. - The group says it accepts no funding from major AI developers. - The full brief and amicus program are available here.

Between the lines: - The filing is a direct challenge to the logic behind early AI fair use wins, not necessarily to fair use itself. - The group is trying to shift the debate from whether AI training is useful to whether existing copyright doctrine fits a machine process at all. - The Human Baseline Framework suggests AI disputes may expose legal assumptions built around human behavior, not automated systems.

What's next: - Judge Breyer will decide whether to allow the amicus filing in the Mosaic LLM litigation. - The broader legal fight over whether AI training fits within fair use is likely to continue in court unless Congress intervenes. - The AI Rights Project says it will use its amicus program in other AI cases.

The bottom line: - The filing asks courts to stop treating AI training and human reading as the same thing until someone proves they are.