Google tests AI-rewritten news headlines inside Discover, but the system misinterprets stories, introduces errors, and frustrates publishers. The experiment reveals rising tension between automation, accuracy, and media responsibility.