Facts
Mark Walters is a nationally syndicated radio host and gun-rights commentator. On 3 May 2023 a journalist asked ChatGPT to summarise a complaint the Second Amendment Foundation had filed that day against Washington's Attorney General. ChatGPT accurately summarised pasted passages, but given a link to the complaint it first said it could not access the internet and then produced inaccurate summaries, the second naming Walters as the Foundation's treasurer accused of embezzlement. Walters was never a party to that case. The journalist had seen warnings that ChatGPT may produce inaccurate information about people and established within about ninety minutes that the output was false. Walters sued OpenAI for defamation in June 2023.
Question
The court had to decide whether a chatbot's false statement about a real person can found a defamation claim against its developer under Georgia law. Three questions arose: whether a reasonable reader in the journalist's position would have understood the output as stating actual facts; what standard of fault applies to an AI developer and whether Walters could meet it, including the actual malice standard if he was a public figure; and whether he could recover any damages, given that he claimed no actual loss and had not asked OpenAI for a correction or retraction.
Decision
Judge Tracie Cason granted OpenAI summary judgment on three independent grounds. Given the disclaimers, ChatGPT's own statement that it could not open the link and the journalist's knowledge that the claim was untrue, a reasonable reader in his position would not have taken the output as stating actual facts, so it was not defamatory as a matter of law. Walters offered no evidence that OpenAI acted with at least ordinary negligence; the court relied on expert evidence of OpenAI's efforts to reduce mistaken output and rejected the idea that releasing an error-prone tool is negligent in itself. As a public figure he also could not show actual malice. He conceded no actual damages, and presumed and punitive damages were unavailable.
Why it matters
The ruling decides, on a full evidentiary record, a defamation claim based on a generative AI hallucination. The court gave weight to the context in which the output was received, including disclaimers and the user's own knowledge, and declined to treat releasing a system known to make mistakes as negligence in itself. As a state trial court order it binds no other court, and its reasoning rests on specific facts, such as an experienced user who did not believe the output; cases where false output reaches a wider audience or causes actual harm may be decided differently.
Related stages
The same court had denied OpenAI's motion to dismiss in January 2024. The May 2025 order is a proposed order submitted by OpenAI's counsel, “as edited by the Court”. The sources reviewed, including commentary from July 2026, report no appeal; the Georgia Court of Appeals docket could not be accessed, so this has not been confirmed from the official record. For a different outcome on different facts, see nanjing-baidu-ai-answer-defamation, where a Chinese court held Baidu liable for an AI answer stating that a lawyer had been jailed.