Navigating if AI is for or against creatives
Be it sword, pen, or AI — can a tool be a villain, or is it how we use it?
AI in creative work is a moral minefield, and creators are right to be wary, angry, and to object.
Companies are using AI for blatant plagiarism, copyright infringement, unauthorized replication of a writer's voice, and disregard for the people who made the work in the first place. That is one way to build, train, and use AI. It is only one.
My view: AI is a tool, and tools themselves aren't villains. But like any tool, it can be misused.
Applied ethically — respecting the IP of creators, leveraging the native strengths of language models, and never training on the creators' work — the technology has just as much potential to help creatives.
Can AI help my mother?
I got my 64-year-old mother into The Wandering Inn, and she absolutely loved it — but, like many readers facing a 15-million-word-and-counting fantasy epic, she was getting gloriously lost in the sauce. It's like drinking from a fabulous fantasy firehose. Characters, places, species, items, groups, classes, levels, relationships, events, timelines, the whole caravan of chaos. So I started building Noveletta: a spoiler-safe, chapter-aware knowledge base to help her track the story as it unfolds.
I cherish stories and world-building, and all the lovely little details authors weave to create them. The challenge in building something to follow them, chapter by chapter, is the sheer tidal wave of details. Each chapter contains hundreds of them, tying together characters, places, events. After evaluating all the options, I determined this is a perfect application of large language models: pulling knowledge out of long-form text is what they're genuinely good at.
But here's the dilemma: authors, like many creatives, are being extorted by large AI companies en masse. So I needed to answer one question before writing a line of it.
If AI touches creative work at all, can this be done without becoming part of the problem?
Creatives' concerns are not hypothetical. The broader AI landscape has given them every reason to be skeptical, guarded, and furious. If I'm going to build something with literature, then I have a responsibility to be extremely thoughtful about where the line is, what the risks are, and what guardrails are non-negotiable.
AI extortion of creatives
The state of AI towards creatives is currently, rightfully, upsetting: blatant plagiarism, copyright infringement, unauthorized voice replication, and disregard for the original creatives — not to mention a bastardization of the craft of art and writing in many cases. This frustrates me, and it is a completely thoughtless application of the technology.
You've probably seen it in the trends where millions of people generate anime portraits of themselves in a style running parallel to a beloved studio's. Other examples look like this:
Training AI models on massive bodies of creative work without meaningful consent, compensation, or even basic respect for the people who made it.
Tools built specifically to imitate the tone, style, or voice of living writers and artists — turning years of craft into a button.
Generated books, illustrations, articles, and media designed to compete with the very people whose work made those systems possible.
The unique intellectual property of artists and their creations is the cornerstone of their career — a career that is near impossible to get off the ground, which is how we ended up with "starving artist" as a cultural idiom. Abusing AI to take that IP does even more damage: it cheapens the perceived value of original work, erodes trust, muddies authorship, and sends the message that the labor of making something beautiful is secondary to the scale economics of reproducing its shape.
That's rot. Artists and authors are right to be mad. Now the question is: who is accountable — or is AI as a technology to blame?
Is AI inherently evil, or is it the wielder?
Tools are just that tools, language itself is a tool and by in large they have no leaning but are the instruments for their users' wishes for example pick one up and turn it over:
Defends a village.
Conquers one.
Same object, both times. The only variable that moved was the hand holding it.
Language models can be used to extort creators' work like training to generate parallel content, which is sickening or to extend accessibility, protect the elderly from cyberattacks, democratize access to education. Now, I'm in no illusion. Noveletta isn't saving the world, nor, frankly, is it fully an AI product. I'm using models to ethically pull structured knowledge, so readers can find their way deeper into literature they love with something a traditional wiki could never do.
That distinction matters. I am extracting knowledge from each chapter as data. I am not imitating the author, pirating their work, or replicating their content. I'm trying to immerse readers in their worlds by helping them follow along with a wiki that doesn't spoil the ending.
The common room had gone quiet again, and the innkeeper set down the glass she had been drying. Out past the road, something in the tall grass was moving, and it was not a person. She reached for the frying pan without looking at it.
That said, I am using AI on source materials, which could be misused or inadvertently expose the author's hard work if I'm not careful. If their material were retained by the foundation models I use, I'd be handing it to training, replication, or other misuse — even when my intent is to expand the reach of their work. So how do I prevent that without becoming part of the problem I hate?
Guardrails for ethical AI use
Because of the current state of AI and its relationship with creative work, I've sweated a lot to be extremely deliberate about the ethical boundaries here. I've read through the contractual protections of every foundation major model as of the writing this, board ethics discussions from industry leaders, and sought feedback from authors, artists, and creatives. What came out is a set of minimum guardrails required to work with creatives at all. Where possible, they're built into the code structurally, not written on a wall or in forgiving terms and conditions.
Open one to see how it's held in place — then try switching it off.
Switching one off here is a demo. In the product there is no switch, it's the only way — that's the entire point of a guardrail.
It's the high transparent road or no road at all
Noveletta is not feasible without AI, and I would rather take it down than cross a boundary set by the creators whose work I treasure. That isn't a hedge; it's the last guardrail, and it's the one that makes the other five mean something.
I will continue to publish how I do this work, the progress I make, the feedback I get so that I am accountable and ethical in the creation of this tool. I am pleased that the authors I'd communicated this to have been open and understanding and approved me to continue to experiment so we can together define what this looks like as we learn it's capabilities.
These guardrails, paired with strong contractual legal protections for the author, are my framework for using AI to support creatives instead of extorting them. A sword, a pen, an language model — none of them decide anything. We do. I'd like the record to show which way I chose.
Noveletta · a spoiler-safe knowledge base for web serials · built solo, on open seas.