Who refused, and what they refused
The European Writers' Council (EWC) and the European Council of Literary Translators' Associations (CEATL) published a joint statement on 25 October. Between them, the two bodies say they represent roughly 272,000 writers and literary translators in Europe.
The document they declined to co-sign is the European AI Office's Code of Practice on transparency of AI-generated content. It is voluntary. By 31 July 2026, about 190 organisations had put their names to it, according to the signatory list published by the European AI Office. The EWC and CEATL are not among them.
The timing matters. The transparency obligations themselves took effect on 2 August 2026, and the Commission's Article 50 guidelines set out how it reads them. Two and a half months later, the book sector's largest representative bodies said the resulting framework is not good enough to put their names on.
What Article 50 asks for, in ordinary language
Since 2 August, anyone offering an AI system in the EU has to make three things clear. People must be told when they are interacting with a machine rather than a person. Synthetic images, audio, video and text have to be labelled as AI-generated. And the content has to carry machine-readable marking that automated systems can detect.
That last part is the technical half of the promise. A label on a cover can be removed by anyone who copies the file. Marking embedded in the data is meant to survive the trip from one platform to the next.
Where the exemptions begin
The two organisations argue that the Code's exemptions and loopholes are wide enough to make it technically useless for the book trade. Their phrase is that exemptions should not swallow the rule. Responsibility, they say, can currently be passed down the chain until nobody holds it.
What they want instead is what the statement calls persistent transparency: unbroken, machine-detectable marking plus human-readable labelling at every step, from the AI provider through the user, the publisher, the distributor, the platform and finally the reader. No single link gets to drop the label.
The human review question
One passage in the statement deserves attention from anyone who works with text. Editing an AI draft, the two bodies argue, does not turn that draft into a human creation. Their example is translation. Someone who revises a machine-produced translation should be credited as an editor, not as a translator, and the fact that a machine generated the underlying text should appear in the metadata and on the copyright page.
This cuts against a common practice in publishing workflows. A tool drafts, a freelancer cleans up, the result appears under a human name. The EWC and CEATL are saying that the cleanup is real work and should be paid and credited as such, but it is a different job from writing or translating from scratch.
Copyright is not in the Code
The statement also criticises the Commission for leaving copyright and authors' rights out of the Code of Practice altogether. Transparency about how a text was produced and the question of who owns the result are two different legal tracks, but for a translator deciding whether to accept a machine-assisted contract, they arrive in the same email.
The AI Act's transparency duties deal with disclosure. They do not settle authorship or licensing. Those remain national copyright law plus whatever contract is on the table, and the statement notes the gap without proposing a fix inside the Code.
What still happens on 2 December 2026
Refusing to sign a voluntary code does not switch off the law. Under Article 50(2), generative AI systems that were already on the market before 2 August 2026 must incorporate machine-readable marking by 2 December 2026. That deadline stands regardless of who co-signed the Code.
It is worth keeping this separate from the delay that got more headlines. The EU's Digital Omnibus on AI, Regulation (EU) 2026/1744, pushed the high-risk obligations for Annex III systems to 2 December 2027, and to 2 August 2028 for AI built into regulated products. Those are the high-risk rules. The Article 50 transparency duties took effect on schedule.
What this looks like for a small publisher
The practical question for a small press in Prague, Warsaw or Lisbon is not whether to sign anything. It is what to write in the fields it already fills in: the metadata attached to an ebook, the credits page, the author bio. The EWC and CEATL are asking for those fields to record machine generation explicitly, and for the record to travel with the file rather than stop at the first upload.
Nothing in the statement sets out a template, and neither organisation has published technical specifications. The demand is addressed to the European AI Office, which finalised the Code in July. Whether the guidelines are reopened is not stated in the document.
Does refusing to sign the Code of Practice mean these organisations are outside the law?
No. The Code is a voluntary instrument. The legal obligations come from Article 50 of the AI Act, which applies to providers and deployers of AI systems in the EU whether or not they signed up to the Code.
If I rewrite AI output heavily, do I still have to disclose it?
According to the EWC and CEATL, yes. Their position is that editing a machine-generated draft does not make the text human-created, and that the AI generation should be disclosed in metadata and on the copyright page. Their example is a revised AI translation, where the reviser would be credited as an editor rather than a translator.
Does Article 50 marking apply to a self-published author?
The transparency duties fall on providers and deployers of AI systems operating in the EU, and the label on synthetic content is meant to follow the content itself. Where a self-published author sits in that chain is one of the points the two organisations say the Code leaves unclear. The Commission's Article 50 guidelines are the document to check for the current reading.
The Code remains what it was on 31 July: a voluntary commitment. The EWC and CEATL, representing some 272,000 writers and translators, are not among its signatories, and their statement does not say what they will do next.
