The Macmillan AI policy has become a practical contract issue for authors preparing proposals, negotiating offers, or asking an agent what protections should appear in writing. Recent reporting says Macmillan Publishers rolled out an internal AI policy framework in July 2026 and that CEO Don Weisberg has said the company’s existing policy forbids using manuscripts to train AI without permission. He also said Macmillan was working on contract language that would make that point explicit for authors who want it written into their agreements, according to PublishingPost.
For authors, the useful question is not whether a publisher has made a public statement about AI. The more useful question is whether the author’s own contract says what will happen to the manuscript, who may access it, whether third-party systems may retain it, and whether the publisher may grant AI-related uses to another company. A policy can guide conduct inside a company. A signed contract is the document authors and agents usually rely on when rights, permissions, and remedies need to be identified. This article is editorial information, not legal advice; authors should ask a qualified publishing attorney or experienced agent before signing or revising contract language.
What The Macmillan AI Policy Says Now
Macmillan AI Policy And Permission
The central point reported in July 2026 was that Macmillan’s policy did not allow authors’ manuscripts to be used to train AI models without permission. That statement matters because manuscript text is not just a production file. It is the unpublished expression an author is submitting for editorial, commercial, and publication use under specific terms.
Still, several authors reportedly asked for the protection to be spelled out directly in their contracts. That request is sensible because a policy can change, may be interpreted differently across departments, and may not answer every practical question that arises during editing, marketing, archiving, vendor review, or future licensing. A clause does not solve every issue, but it can reduce ambiguity by naming the permitted and prohibited uses.
Education Rules Are Not Trade Contract Terms
Macmillan-branded education policies should not be confused with Macmillan Publishers trade-book contract terms. Macmillan Education, part of Springer Nature, published an AI Editorial Policy updated on December 17, 2024. That policy said AI tools cannot be credited as authors, generative AI images are not allowed in publications, and authors must disclose the use of generative AI tools in an introduction or acknowledgments, according to the Macmillan Education policy.
Those education-policy points are relevant because they show one formal approach to AI disclosure and authorship. They do not, by themselves, tell a novelist, memoirist, journalist, academic trade author, or children’s writer what language will appear in a Macmillan Publishers trade contract. Authors should keep those categories separate: editorial disclosure rules, publisher operating policies, and author contract clauses each do different work.
Why Authors Are Asking For Contract Language
Policy Is Not The Same As A Clause
The Macmillan AI policy discussion has sharpened a wider author concern: whether good intentions are enough if the contract is silent. In proposal and submission work, silence can create stress long before a manuscript reaches copyediting. A nonfiction proposal may include sample chapters, chapter summaries, platform material, research notes, and marketing positioning. A fiction submission may include the full manuscript, synopsis, and related pitch materials. Authors increasingly want to know whether those materials may be entered into AI-assisted systems, whether such systems retain data, and whether the publisher can authorize a vendor to process the text.
Recent reporting described several contract requests that authors and agents have raised with Macmillan. These included an explicit exclusion preventing manuscript text from being used for AI training; permission for limited internal tools such as spell-check or grammar assistance only if they do not retain author data; restrictions on sublicensing AI-training rights to third parties; and a reversion concept if Macmillan were acquired by a company whose main business involved AI development. The availability and acceptance of any clause can vary by book, bargaining position, agent strategy, imprint, and timing.
Four Questions To Raise Before Submission
Authors do not need to wait until a final contract draft appears before organizing their questions. Clear questions can help an agent, editor, or publishing attorney identify the gap between policy and enforceable language. They can also keep proposal conversations focused on rights rather than vague reassurance.
- Will the contract state that the author’s manuscript and proposal materials cannot be used to train internal or external AI models without express permission?
- Will any AI-assisted editorial, production, accessibility, marketing, or analytics tools be limited to systems that do not retain or reuse author text for model training?
- Will the publisher be barred from sublicensing AI-training rights to vendors, affiliates, or future partners unless the author gives separate written consent?
- If ownership of the publisher changes, will the author have any defined protection if a buyer’s main business involves AI model development?
These questions are not demands every publisher will accept. They are screening tools. If an author is preparing a proposal, the answers may shape how much material is submitted, what confidentiality expectations are discussed, and which contract points the agent flags early.
Submission And Proposal Steps Authors Can Use
For Agented Authors
An agented author should ask the agent what AI language the agency is currently seeking and whether the agency has seen different responses across imprints. The agent may know whether a publisher’s boilerplate has shifted, whether certain protections are available only by request, and whether a specific editor has authority to discuss nonstandard terms. Authors should avoid assuming that a public policy automatically appears in the contract draft.
A practical approach is to create a short rights memo before the proposal goes out. That memo can list the author’s priorities: no AI training without permission, no third-party retention of manuscript data, no sublicensing of AI-training rights, disclosure of any AI-assisted editorial tools, and confirmation that standard editing or production tools do not expand the publisher’s rights. Authors looking for clause-level thinking may also want to compare related discussion of AI consent clauses before discussing terms with an agent or attorney.
For Unagented Authors
Unagented authors face a different problem: they may receive a contract without an advocate already in the conversation. That does not mean they should sign quickly. If an offer arrives, the author can ask the publisher whether its contract includes explicit AI-training language and whether manuscript files will be processed by any external AI systems. If the answer is broad or unclear, the author can request written clarification before signing.
The Macmillan AI policy debate also shows why proposal-stage records matter. Authors should keep copies of submission guidelines, email exchanges, offer letters, contract drafts, and any written explanations of AI use. Those records may help an attorney or adviser understand what was promised, what was only described informally, and what actually entered the signed agreement.
- Save the exact contract draft and every redlined version.
- Keep publisher emails that describe AI, data retention, vendor use, or training restrictions.
- Ask whether protections apply to proposals, sample chapters, full manuscripts, edited files, audiobook scripts, metadata, and marketing copy.
- Separate creative-use consent from routine production tools so the discussion does not become too broad to manage.
For those researching their options, if you’re exploring various resources within this site network, Bethel NC UMC is another associated website, but it should not be used as a source for interpreting publishing contracts. Such legal queries should always be addressed directly through the contract language and with the guidance of qualified advisers.
Macmillan AI Policy For Submission Planning
The Macmillan AI policy issue is best handled before an author is under pressure to sign. For proposal writers, that means treating AI rights as part of the submission plan, not as a late-stage technicality. A proposal can be polished, market-aware, and persuasive while the author still asks reasonable questions about how the text will be handled.
One practical distinction can keep the conversation grounded: not every AI-related use is the same. A spelling tool that does not retain author data raises different concerns from a system that ingests manuscripts to train or improve a model. A publisher’s internal workflow tool raises different questions from a third-party vendor. A one-time editorial support process raises different issues from sublicensing author text for future model development. Contracts work better when they separate these categories instead of referring to “AI” as one single activity.
Authors should also ask who can approve exceptions. If a contract says AI training requires permission, the author should know whether that permission must be separate, written, and specific to the proposed use. Broad language can weaken the point of consent. Clear language can say what material is covered, which party may use it, for what purpose, for how long, and whether compensation or withdrawal rights apply. Those are legal drafting questions, so the safest course is to have a publishing attorney or experienced agent review them.
The reporting available as of October 6, 2026, suggests movement but not full uniformity. Macmillan’s leadership said manuscripts were not to be used for AI training without permission and that explicit contract language was being developed for authors who wanted it. Some authors still wanted stronger, written protections. That tension is the practical lesson: a publisher policy can be reassuring, but proposal-stage authors should ask whether the protection they care about appears in the signed contract.
For authors preparing submissions now, the Macmillan AI policy is a prompt to be orderly. Finish the proposal. Track the materials you send. Ask direct questions about AI training, retention, vendors, and sublicensing. Get answers in writing where possible. Then make contract decisions with professional advice rather than relying on informal assumptions.