.png)
Your Unpublished Manuscript and AI: Do You Know Where Your Words Go?
There is a new AI authorship question writers should be paying attention to, and it is much closer to home than the familiar copyright lawsuits over published books: what happens to work that has never been published at all?
This week, mathematicians raised concerns that private or unpublished research shared through AI tools may have contributed to later model capabilities or high-profile results. The claims are disputed and have not been proven, but the argument matters well beyond mathematics. Writers routinely paste outlines, sample chapters, character notes, research, synopses and sometimes entire manuscripts into AI systems. That makes the question immediately relevant to authors.
Quick Shortcuts
The story has moved from piracy to provenance
The Verge reported on 11 September that mathematicians are asking for greater transparency about whether their conversations and unpublished work were used in model development. Read The Verge report. The important point for writers is not to assume those allegations are established fact. It is to notice the wider issue they expose: provenance is becoming part of the AI trust debate.
For the last few years, authors have understandably focused on whether published books were scraped, copied or pirated for training. That fight is still very much alive. But writers now need a second layer of awareness: what are we voluntarily handing to AI systems ourselves?
The settings matter more than most writers realise
OpenAI currently states that content from individual services such as ChatGPT may be used to improve models unless the user opts out. It also says business products and the API are not used for training by default. OpenAI data-use policy The company also provides Data Controls that allow users to turn off use of new conversations for model improvement. Data Controls guidance.
That distinction matters. “Using AI” is not a single activity. Asking for ten alternative headlines is very different from uploading an unpublished novel. Generating an advertising concept is different from feeding in a complete world bible containing years of original development. The risk, value and sensitivity of the material are not the same.
AI-assisted is still not AI-authored
My own position remains consistent. I write my books. AI can help me research, visualise, organise, market and operate the creative business surrounding those books. Those things are useful precisely because they leave the author at the centre of the work.
But that does not mean every piece of unpublished creative material should be treated casually. An author can be enthusiastic about AI as a tool and still be careful about where a draft manuscript goes, which product is being used, and what data controls apply. There is no contradiction there. It is simply good digital housekeeping for creative work.
A practical rule for authors
Before putting unpublished material into any AI system, ask three questions: Do I understand whether this service may use my input for model improvement? Have I checked the relevant privacy or data controls? And do I actually need to provide the whole manuscript to accomplish the task?
Often the safest and most useful approach is also the simplest: provide only the material needed for the job. A paragraph may be enough to test tone. A scene summary may be enough to brainstorm marketing copy. A character description may be enough to create a visual reference. You do not automatically need to surrender the entire creative archive just because the tool accepts large files.
Publishing now has a trust problem on both sides
Writers are increasingly expected to be transparent about how AI touched their work. That is reasonable. But transparency should run in both directions. If authors are asked to disclose how AI was used in research, editing or production, AI companies should be able to explain clearly how user-supplied creative material is handled, retained and used.
This is where the debate is heading. Copyright asks who owns the work. Provenance asks where the data came from. Privacy asks what happens when you hand over something that was never public in the first place. For working writers, all three are now part of the same conversation.
The question for writers
Would you put an unpublished manuscript into an AI system today and do you know exactly what happens to it after you press send?
About Rob Frankson
Rob Frankson is a science-fiction author and creator of the Near Galaxy Saga, writing about authorship, publishing, technology and the practical realities of using AI around a human-led creative process. More about Rob
AI & Editorial Transparency
AI & The Author is edited and published by Rob Frankson. Artificial intelligence is used to assist with news research, initial drafting, content organisation and supporting imagery. All articles are reviewed and, where necessary, edited by Rob Frankson before publication. The opinions, editorial position and final decision to publish remain the author's.
Comments