Not going to lie; it’s a bad idea
Without getting into the moral or environmental issues of using generative artificial intelligence (GenAI) – specifically Large Language Models (LLMs) such as Claude, GPTs and Gemini – here are some basic facts you should be aware of if you are considering using AI to assist with writing your manuscript.
First, let's examine what AI is
The progression of assistive software to traditional AI and GenAI
To keep things simple, we’ll compare these tools when applied to writing only.
Traditional software
This falls into the category of Grammarly or Word’s original spelling and grammar checker. This software was coded to create a basic database with a strict set of rules. The output always produced the same programmed response. The software did not grow, learn or compare, although an author could add to the database or create their own dictionary.
Traditional AI
Along came a smarter facility that used machine learning and natural language processing. This meant spelling and grammar checkers could now analyse context and offer varying suggestions, rather than rigid corrections.
Enter the behemoth, GenAI
GenAI is a completely different animal to previous technology. Rather than referring to a fixed database, it creates new content via LLMs, which reference massive data sources that are continually growing by scraping online resources. This technology has now been integrated into most writing packages, including Word’s Copilot, whether authors want it or not. It can, however, be turned off: Turn off Copilot.
The issue we now face, apart from GenAI being unreliable due to its known hallucinations, is that its generation of new content – whether a sentence, paragraph or a whole chapter – means GenAI is effectively a co-writer. And this is where a major problem arises: copyright.
Why copyright is now an issue
Ownership
If an author allows GenAI to generate content for them, even if it’s accepting a suggested rewrite of a paragraph rather than a whole chapter, the author no longer owns copyright of the entire work.
Why does ownership matter?
This means that when the author moves to publish their work:
- They cannot enter writing competitions.
- Traditional publishers will not accept manuscripts where AI has been used.
- Traditional publishers now require authors to provide evidence of how their work has been developed over time – this may include earlier drafts, notes and research.
- Self-publishing platforms require authors to declare use of AI, else they risk having their account shut down. Amazon has already done this to some established authors.
Author contracts cancelled due to suspected AI use
Here are two examples of authors who made headlines for losing their publishing deals:
Shy Girl by Mia Ballard – article by The Conversation
Call Me, I’ll Hide the Body by Jerry Falade – article by The Guardian
New embedded watermarks required by EU laws
If you think you can get away with simply saying you haven’t used GenAI when you have, you should be aware that LLMs are now required by EU law to carry invisible watermarks, which cannot be removed. This new technology will carry across all versions of LLMS released round the world.
There has been a lot in the news recently about software platforms like Pangram, which detect use of AI in writing; however, it’s commonly thought they produce false positives. The new watermarks operate differently, and are embedded directly into the words, so there’s no escaping them.
But the work is all my ideas, plot points and characters!
If you are under the impression that you can simply claim the ideas are yours and are therefore copyright, there’s bad news: ideas are not copyrightable; it’s only the words you write down that matter – the physical expression of the ideas.
Honestly, while plot points and characters sit at the core of a novel, they don’t have a huge bearing here, as they provide little that is new – every story has been told over time.
It’s more about how the story is told – the expression and execution of the work. That’s what makes it unique – your own words written in your own distinct voice and style. This includes the phrases, similes and metaphors you use, the structure of your sentences and the unfolding of paragraphs and scenes that immerse the reader and hold their attention through generating genuine human emotion.
And if those words are generated by AI – even a portion of them – you don’t own that work.
Read more about the perils of AI
Part 2: Why AI writing has no voice: GenAI Risks Part 2
This blog covers why GenAI is so bad at writing long-form writing and how readers and editors spot it so easily.
Part 3: What GenAI “tells” look like: GenAI Risks Part 3 (coming soon)
This blog provides a list of examples that flag typical AI writing.
