Big Five Publishers' Undisclosed AI Use

The Big Five publishers, including HarperCollins, Simon & Schuster, and Hachette, are currently suing AI companies over alleged copyright infringement from training large language models on their books ^[1]. At the same time, internal sources describe those same publishers quietly routing cover art, publicity copy, and agent correspondence through AI tools, with none of the three disclosing to readers or authors which covers, blurbs, or pitches involved AI ^[2]. The contradiction is structural, not incidental: the industry's public posture treats AI training on books as theft, while its private production line treats the same technology as a labor-saving utility.

HarperCollins, Simon & Schuster, and Hachette arrived at internal AI adoption through distinct paths. HarperCollins bought Claude licenses from Anthropic and designated dozens of staff as 'AI Champions' tasked with surfacing new use cases in monthly brainstorming sessions, while also paying for ChatGPT and the marketing platform Jasper; one division reportedly went 'all in,' including AI-generated marketing videos, even as its legal team told editors no when they asked about pasting full manuscripts into AI tools ^[2]. Simon & Schuster took a louder, more top-down route: OpenAI representatives ran an in-house workshop suggesting ChatGPT for composing emails to literary agents, and the publisher ran an internal contest with a $10,000 top prize for AI use cases before easing the push after employee complaints ^[3]. Hachette, by contrast, draws an official line, stating it supports AI tools for internal operational purposes but not for creative uses ^[4].

Employees across all three publishers describe the same underlying pressure: layoffs and cost-cutting have doubled or tripled individual workloads, making AI shortcuts hard to resist even for staff who find the practice distasteful ^[3]. Publishing commentator Kathleen Schmidt argues the industry's 'do more with less' culture is the real driver of internal AI adoption, calling the use of AI to generate publicity copy 'short-sighted' ^[4], while University of Texas Press publicist Joel Pinckney faulted leadership for adopting the tools out of competitive fear rather than strategy, saying 'the lack of any kind of vision from leadership beyond the logic of inevitability and the fear of being left behind is breathtaking' ^[5]. On Reddit, self-identified publishing employees described a split reality inside the buildings themselves - internal AI tools for meeting transcription and repetitive-task automation are embraced even as cover art remains an explicit red line, with legal approval required for AI-related questions and companies maintaining their own internal AI systems specifically to avoid uploading sensitive work to outside tools. On X, one author under contract described having to personally rewrite an 'LLM-garbled' back-cover description for her own book, adding pointedly that her publisher was happy to profit from AI licensing deals even while rolling the technology out quietly in-house. YouTube commentary pointed to a more public version of the same shift: a HarperCollins job listing for a Creative Director role explicitly required fluency with AI design tools, suggesting the quiet internal rollout is starting to surface in open recruiting as well.

The response has moved from grumbling to contract language. In April 2026, the Authors Guild issued a statement and introduced new model contract clauses barring publishers from uploading manuscripts or authors' personal data into consumer-facing AI systems without written permission, amid concerns that editors could do exactly that ^[6]. Literary agents followed with their own clauses stating flatly that 'this material can't be run through an LLM' ^[4]- language that, by its plain wording, also restrains the publishers themselves from feeding client work into their own in-house AI tools. Some agents have gone further, comparing rejection letters across different editors and spotting identical stock phrasing and sentence constructions that give away machine-generated prose ^[2]. Author Kelly Link said the controversy has already damaged relationships between writers and the publishers who represent them, a sign the trust cost may outlast any productivity gain ^[5].