A writers' guild spells out what an AI clause should say

The process wall

A debut author signing a standard trade contract typically accepts boilerplate that predates generative AI entirely; without a specific clause, that silence can read as permission for a publisher to use a work in ways the author never separately agreed to. The Authors Guild's own advocacy page states it added a clause to its Model Trade Book Contract in March 2023 barring the use of a work ‘for training artificial intelligence technologies without the author's express permission,’ then introduced ‘four new clauses concerning AI’ that June. The wall the Guild addressed is exactly that gap between silence and consent.

What the documents show

The Guild's model clauses page gives the clause language directly. A training clause states the ‘Publisher acquires no rights to … use or create derivative works of the Work … for purposes of training generative artificial intelligence technologies’ unless expressly negotiated. A narration clause bars an audiobook being produced with ‘AI or other non-human narrator’ without the author's ‘prior and express written consent,’ and a parallel clause requires the same consent before an AI translation. A further clause bars a publisher from uploading ‘the Work or any of Author's personal information to consumer-facing AI systems … without written permission.’ Separately, the Guild's best-practices guidance tells authors to ‘publicly disclose any substantial use of AI to generate text,’ reasoning that ‘readers have the right to know whether the books they buy, borrow, and read were written by a human.’

Authorship and limitation

The mechanism worth naming: a model clause is a negotiated contract term, not an automatic right. It binds only the specific publisher who signs it, and only if an author or agent actually inserts it into a given deal; the Guild cannot compel any publisher to adopt its language. The disclosure guidance is a separate obligation aimed at readers and at the Copyright Office rather than at a publisher: the same guidance warns authors to ‘disclaim any AI-generated text … on your copyright registration application’ to avoid a fraud finding, which is a distinct legal exposure from anything a contract clause covers.

What survives

An author negotiating a contract can ask, clause by clause, whether training, translation and narration rights are reserved or granted, since standard boilerplate may say nothing about any of them. Disclosure to readers is a step an author controls regardless of contract language, and the guidance suggests doing it ‘by describing the AI use in the front material of a book or at the bottom of an article’ — an editorial suggestion about placement, not a rule any contract enforces.

• Ask directly whether a contract's standard language is silent on AI training, translation, or narration rights before signing.
• Decide where a disclosure of substantial AI use belongs in a specific book — front matter or a closing note — before publication.
• Confirm any AI-generated text is disclaimed on a copyright registration application before filing it.

The guidance gives an author language to propose and a reason to disclose, but neither exists automatically in a contract or a registration until the author or their representative puts it there.