The publishers allege ChatGPT scraped nearly 100,000 copyrighted articles to train its models and reproduces their content verbatim in user responses—while attributing hallucinated text to their ...