Major US book publishers are publicly fighting generative AI in court while quietly embedding it into their own workflows. According to interviews with over two dozen anonymous industry workers, HarperCollins, Simon & Schuster, and Hachette have been using LLMs like Claude and ChatGPT for tasks ranging from drafting emails to literary agents to generating back-cover copy and marketing materials. These practices have proceeded without author consent or formal public disclosure, even as the same companies pursue copyright lawsuits against AI developers.
Employee reactions have been largely negative. At HarperCollins, staff were assigned to become "AI Champions" and tasked with finding use cases in monthly brainstorming sessions. When concerns about legal, ethical, and environmental impacts were raised, a meeting leader reportedly dismissed them as "the cost of doing business." At Simon & Schuster, a $10,000 AI contest was later rolled back after repeated employee complaints in town halls. One publicist called AI-generated publicity copy "short-sighted," while an independent publisher noted that such efficiency-driven adoption feels "pretty inevitable" for large corporate houses.
The gap between public posture and internal practice is stark. Publishers have sued Google and canceled books by authors suspected of using LLMs, yet the same firms are buying enterprise licenses and experimenting with AI for operational tasks. A Hachette spokesperson said the company supports AI for operational purposes but not for creative uses or author communication—a line that workers say is already being blurred in practice. The single source for this article is a WIRED investigation; no other independent reporting was provided for comparison, so all claims rest on that account alone.