An authors' lawsuit put AI training itself on trial

The writer's problem

On 19 September 2023, seventeen novelists and the Authors Guild filed a class-action complaint against OpenAI in the Southern District of New York, as the guild's own announcement of the filing shows. The named plaintiffs, including George R.R. Martin, John Grisham and Jodi Picoult, brought the suit on behalf of a proposed class of fiction writers whose books they said had been used to train GPT-3.5 and GPT-4 without permission or payment. The problem was not a disputed chatbot answer; it was the training step itself, which the writers said had already happened to their published work before any prompt was typed.

What the documents show

The announcement states the guild's council and foundation board voted, with abstentions, to authorize the suit, and that the complaint alleges the plaintiffs' books were downloaded from pirate ebook repositories before being copied into the models. On 4 December 2023 the plaintiffs filed an amended complaint naming Microsoft as a co-defendant, extending the same allegations to OpenAI's principal financial backer and distribution partner. These remain allegations pleaded by the plaintiffs; the guild's own materials describe what the complaint says OpenAI did, not a judicial finding that it did so unlawfully. The case was later folded into a consolidated proceeding: the judicial panel's own transfer order lists this action, docketed as 1:23-cv-08292, among twelve suits centralized before one judge in 2025.

The editorial choice

For an author approached about joining such litigation, the guild's account frames the choice as collective rather than individual: sign onto a class claim built around shared allegations of piracy and uncompensated training, or pursue a separate licensing conversation instead. That is an editorial framing, not a legal instruction, and the guild's own page is explicit that it is advocating a position, not reporting a neutral outcome. Editors and publishers watching the case had to decide whether to wait for a ruling before revising contract language on AI rights or to act on the complaint's allegations immediately.

What stays with the author

Whatever the suit eventually decides about training, it does not resolve who wrote a given sentence or paragraph, and it does not certify any book as free of AI assistance. The named plaintiffs remain the credited authors of their own catalogs regardless of the litigation's outcome, and the complaint's allegations are not themselves a finding that any specific book was reproduced. A court's later view of fair use, discussed in the guild's own account of a 2026 motion in the same case, does not retroactively change what a reader was told about a book's authorship back in 2023.

• Does a report about this case describe an allegation, a settlement, or a ruling?
• Which entity is named as defendant, and did that change as the case was amended?
• What does the cited primary document itself claim, apart from what a news summary says it claims?

The 2023 filing is best read as the opening position in a dispute that took years to reach a ruling, not as a resolved account of what training data any model actually contains.