The cancellation of 'Shy Girl' by Hachette is the publishing industry's equivalent of a canary in the coal mine — and the New York Times investigation that surrounds it is a serious piece of reporting that the industry would be unwise to dismiss as an isolated incident. The core finding is damning in its simplicity: publishers have no reliable way to detect AI-generated fiction, and most have not yet developed the contractual or editorial frameworks to address it. The submission pipeline — already under pressure from the sheer volume of manuscripts that digital tools have made easier to produce — is now absorbing a category of content that is, by definition, designed to be indistinguishable from human writing. The Hachette case is notable because it involves a major house, a contracted title, and a post-acquisition discovery. The question it raises is not merely 'how do we screen submissions?' but 'what happens when AI-assisted writing passes every editorial review and reaches the production stage?' Authors Guild certification programmes and AI-detection tools offer partial answers, but neither is foolproof, and both create new legal and ethical questions about the burden of proof. The deeper issue is one of trust: the relationship between publisher and author has always rested on an implicit understanding that the work being submitted is the author's own. AI does not simply complicate that relationship — it dissolves the premise on which it is built. The industry needs clear contractual language, transparent disclosure policies, and a shared understanding of what 'human-authored' actually means in a world where AI is embedded in every stage of the writing process.