Strip away the rhetoric, and the industry’s position is legible and consistent. It is not an anti-AI position. It is a position about who controls the economic benefits of AI.
At the US Book Show in New York, three publishing CEOs sat before a full house and performed the industry’s now all-too-familiar two-step on artificial intelligence.
Dominique Raccah of Sourcebooks warned against a “witch hunt” atmosphere that might distract the industry from AI’s genuine opportunities. David Shelley of Hachette Book Group spoke of publishers’ “unique function” as curators – the human intelligence that distinguishes publishing from becoming “a kind of tech pipeline.” Madeline McIntosh of Authors Equity declared that every publisher is “five minutes away from having an AI controversy.”
All three, in the same breath, confirmed they are deploying AI across their operations.
This is not a contradiction those executives even appeared to notice. Which is itself is illuminating.
The Selective Labour Theory
Publishers have constructed, with meticulous care, a moral circle around creative labour.
Inside the circle: authors, illustrators, narrators, translators – the human artists whose displacement by AI is framed as an existential crisis, a theft, an assault on imagination itself.
Outside the circle: engineers, warehouse operatives, phone operators, metadata specialists, marketing copywriters, junior editorial assistants – the human workers whose displacement by the same AI is framed as efficiency, innovation, or simply not mentioned at all.
The PW panel captured this with (unintentional) precision. Raccah noted that her team is using AI coding tools to build internal resources “without outside engineers.” Those engineers – real people who previously did that work – have lost it to AI tools built on the same training data publishers are simultaneously litigating over in federal court. They are invisible in the moral accounting.
Shelley described AI as “already transforming Hachette’s distribution centres, where bookstore queries about the status of their orders, which were previously handled by phone, can now be resolved in seconds.” The human beings who previously handled those calls are not mentioned. Their displacement registers only as a customer service improvement.
This is not incidental. It reflects publishing’s long-standing, pre-AI hierarchy: the author as romantic soul at the centre; everyone else as infrastructure.
AI simply forces the hierarchy into the open by requiring publishers to state their values explicitly. And when they do, the values are clear: creative labour that generates copyrightable product is supposedly sacred (except when it comes to contracts and royalties). Operational labour that generates cost is a necessary evil that is increasingly unnecessary.
The Tiered Moral Framework: Who Gets Sued, Who Gets Paid
The industry’s AI ethics do not apply uniformly to AI companies either. They apply in precise correlation with commercial dependency.
Companies publishers can afford to attack receive the full existential-crisis register. Meta is accused of “industrial-scale copyright infringement” and “brazen theft.” Perplexity is a pirate operation. Anthropic’s training on books from shadow libraries produced a $1.5 billion settlement. OpenAI faces consolidated multidistrict litigation in the Southern District of New York involving some of publishing’s most prominent houses.
Companies publishers cannot afford to attack receive quiet accommodation.
Amazon – which owns the world’s dominant book retail platform, the world’s largest audiobook service, the world’s largest self-publishing infrastructure, and through AWS provides the cloud backbone for much of the AI industry itself – is conspicuously absent from publishing’s moral campaigns.
Amazon’s Audible launched its Virtual Voice programme, offering AI-generated audiobook narration at minimal cost, directly threatening the professional narrators whose “existential crisis” publishers claim to champion. Over seventy authors, including Dennis Lehane and Lauren Groff, signed an open letter demanding publisher commitment to human narrators. The response from publishers was continued partnership with Audible. When the villain is also the landlord, the op-eds don’t get written.
Amazon’s KDP platform has been flooded with AI-generated content – fake memoirs, AI-illustrated children’s books with hallucinated artwork, mass-produced non-fiction generated at near-zero cost – since 2023.
Amazon’s response was regulatory, not prohibitive: disclosure requirements, upload throttling, internal quality flags. Every KDP upload, however generated, is a product generating page-read fractions through Kindle Unlimited. The platform’s financial logic runs on volume regardless of origin.
And the publishers who filed amicus briefs defending human creativity have said almost nothing about it, because Amazon’s algorithm is how their own books get found.
News Corp provides the single corporate entity that holds the most contradictions simultaneously. Dow Jones and the New York Post sued Perplexity for copyright infringement. News Corp simultaneously holds a reported $250 million, five-year licensing deal with OpenAI – a deal News Corp CEO Robert Thomson celebrated by praising OpenAI as a “principled company” and declaring he would “rather woo than sue.”
The distinction between the praiseworthy OpenAI and the piratical Perplexity is not philosophical. It is financial. One company paid. The other did not.
News Corp also owns HarperCollins – which is simultaneously a plaintiff in the May 2026 lawsuit against Meta for training Llama on copyrighted books, a partner with ElevenLabs deploying AI voice-cloning to generate audiobooks from its backlist (targeting titles that “wouldn’t pencil out” financially with human narrators, in the trade’s own framing), and the company investing $160 million in a 1.6 million square foot automated logistics facility in Brownsburg, Indiana, designed explicitly to reduce human labour in its distribution operations.
One corporate family. Three positions. No shame.
The Body Count in the Warehouse
The warehouse is where publishing’s AI hypocrisy becomes most concrete – and least visible, because it takes place far from the literary festivals and book fair panels where the industry’s self-image is constructed.
HarperCollins’s Indiana facility, scheduled to open in 2028, will feature goods-to-person automation – a system in which robots bring products to workers rather than workers moving through the facility – alongside algorithmic Labour Management Systems that track workforce productivity in granular, continuous detail. The company’s own communications describe the design philosophy as “minimising the need for workers to move around the facility.” This is, in the most neutral possible terms, a $160 million investment in needing fewer human beings.
The Simon & Schuster distribution centre in Riverside, New Jersey – one of two facilities handling a combined 165 million books a year, with between 40 and 60 tractor trailer trucks leaving daily – has been automated for years.
The results, on record from Vice President of Logistics Operations Dave Schaeffer, are unambiguous: more books shipped without increasing staffing levels, variable labour productivity up by more than 10%, overtime down by more than 20%, and reductions in fixed staffing requirements.
Schaeffer’s own framing of the rationale is worth noting: “It’s a competitive advantage to have a cost-effective distribution network, and our reputation for efficiency and a wide reach in sales and distribution makes us attractive to many authors and external publishing clients.”
Attractive to authors, built on fewer human workers. Is that the sound of fury and outrage from the publishing community I hear? No, don’t be silly. Mere manual labourers do not warrant our sympathy.
Publishers, it seems, are an existential threat to warehouse workers, but who gives a f**k? Not the publishers who are in the business of making money. And not the authors and other creatives ranting against AI while their books are shifted through the aforementioned warehouses by robots. But no, please don’t try say your publisher is different, so you have the moral high ground.
Ingram Content Group – the distribution infrastructure that thousands of publishers, including many now litigating against AI companies, depend upon – deployed AutoStore’s robotic system across its network, with 160 autonomous robots operating across a high-density grid. The savings were estimated at over $187 million over ten years, driven primarily, in their consulting partner’s explicit language, by labour reduction.
Penguin Random House closed its Kirkwood, New York warehouse in 2014-15, laying off 286 workers, consolidating US distribution from four facilities to two. Its Spanish distribution centre, opened around 2025, features 54 robots for small-order fulfilment and 28 self-guided collaborative robots for high-volume picking, with capacity for over 40 million copies annually.
Industry-wide, warehouse automation typically delivers 25-30% reductions in labour costs – which constitute 50-70% of total warehousing budgets, making human beings the primary target of every efficiency investment. The technology does not hide this. The Labour Management Systems publishers are installing are described in industry literature as offering “granular, scalable, and relentless” performance tracking – squeezing more productivity from fewer workers, or providing the documented justification for further headcount reduction.
The warehouse operatives losing work to robots do not have an Authors Guild. They do not have literary agents amplifying their grievances. They do not get panels at BookExpo or op-eds in the New York Times. The Association of American Publishers, whose members are installing these systems, does not file amicus briefs on their behalf. They are, in the industry’s moral geography, outside the circle.
All workers are equal and important. But some workers are more equal and important than others.
The Longer History Publishing Forgets
Publishing’s current posture of defending human labour against technological displacement is historically novel, and selectively applied.
Publishers did not mourn the typesetters displaced by desktop publishing in the 1980s and 1990s. They did not hold panels about the existential crisis facing the reps whose territory sales roles were hollowed out by online ordering systems. They did not publish open letters about the royalty accountants replaced by automated royalty platforms, the rights assistants replaced by rights management software, or the foreign scouts made redundant by digital communication. The postal workers who once moved manuscripts between authors and publishers, the typewriter manufacturers whose products were essential to literary production – none of these displaced human workers generated the moral infrastructure that now surrounds the author.
Publishing’s relationship with technological labour displacement is long and pragmatic. What is new is not the displacement. What is new is that the technology is now approaching the labour that publishers need to produce their product – and that they have built a public moral framework around protecting, because that labour is also the source of their copyrightable asset and their public legitimacy.
This is not hypocrisy as personal failing. It is institutional self-interest operating through the available moral language. The language of protecting human creativity is not false – publishers do, in some genuine sense, value their authors. But the valuation wilts under scrutiny when it is channelled as an attack on AI companies while the same institutions are quietly running $160 million automation projects and partnering with voice-cloning start-ups to eliminate narrator fees.
The Actual Position
Strip away the rhetoric, and the industry’s position is legible and consistent. It is not an anti-AI position. It is a position about who controls the economic benefits of AI.
AI training on publisher content without payment is theft. AI training on publisher content with payment is a licensing opportunity – and the AAP’s own amicus brief in the Meta case noted, apparently without irony, that the AI training licence market is currently valued at $2.5 billion and projected to reach $30 billion within a decade. That figure appeared in a document nominally defending authors.
AI replacing the labour of narrators, translators, warehouse workers and phone operators is efficiency and competitive advantage. AI replacing the labour of authors – which is to say, AI replacing the production of the copyrightable asset that underpins the entire business model – is existential threat.
The News Corp Chief Revenue Officer stated the actual position with admirable clarity when asked about the company’s simultaneous OpenAI deal and Perplexity lawsuit: “If our content is going to be used, we need to be properly paid for it.” Not: AI is wrong. Not: human creativity must be protected. Properly paid.
Some authors agree with that. Most, if the selective news reporting is to be believed, view AI an an abomination and they want no part in it.
Publishers do not have that luxury of principled posturing. They are businesses and they will follow the money. And the money trails leads inexorably to embracing AI whether authors and other creatives like it or not.
That tipping point is fast approaching. Publishers want to be seen as the good guys defending culture and jobs, but the reality is robots and automation. And don’t for one second imagine the paying public will notice or care.
Lee Jarit, Audible’s global head of publisher and partner relations announced this past week that over 50 million minutes of AI generated content have been listened to on Audible. In Jarit’s words, “these are largely niche titles that would never have been in audio before, and we are getting real engagement from customers”.
Back to the US Book Show: The Sourcebooks CEO, at the same USBS panel where she warned against an AI witch hunt, described her team’s use of AI coding tools as “an opportunity for an industry historically staffed by liberal arts graduates.” The engineers those tools replaced were not liberal arts graduates seizing a new opportunity. They were people who had previously been paid to do that work. Who are likely still paying off huge student debts that let them train for those coding jobs.
A Note on Who Gets to Have an Existential Crisis
Publishing’s current AI discourse has a recurring protagonist: the human author, facing displacement by generative tools trained on their own work, their voice cloned, their market flooded with synthetic imitations, their livelihood threatened. This is a real concern, no question, and it deserves serious engagement.
But the discourse has some conspicuous absences.
The audiobook narrator booking fewer sessions because Audible’s Virtual Voice programme is expanding across the backlist. The literary translator whose per-word rates are collapsing as publishers experiment with AI-assisted workflows. The junior marketing assistant whose entry-level role has been absorbed into an AI tool that generates copy and metadata at a fraction of the cost. The warehouse picker whose job was eliminated when the robots arrived, who will not be quoted in Publishers Weekly and will not be invited to testify before the Copyright Office.
These workers are also facing an existential crisis. Theirs is simply less useful to publishers as a rhetorical instrument, because their displacement serves the bottom line rather than threatening it.
The moral circle publishers have drawn around creative labour is not drawn where it is because that is where human dignity is located. It is drawn where it is because that is where the copyright is located.
The warehouse worker has no copyright. The narrator’s voice, once cloned, belongs to the algorithm. The translator’s work, once automated, is a cost centre. The author’s manuscript is the asset – and the asset needs protecting, which is why the author gets the op-ed, the open letter, and the amicus brief.
Publishing is not anti-AI. It is pro-revenue, as all industries are, and AI is currently both a threat to some revenue streams and an opportunity for others. The threat gets the moral language. The opportunity gets the press release.
The hypocrisy itself is unremarkable. What makes it worth examining is how completely the industry has persuaded itself – and much of its public – that the moral language is the whole story.
It isn’t. The warehouse is part of the story. The narrator is part of the story. The engineer is part of the story. The $250 million OpenAI deal signed by the same company suing the AI company that didn’t pay is part of the story.
