Charlie Stross, a Hugo-winning author, refuses to use AI in his writing process, calling it a “grift” that steals from real writers. His novels, including “Singularity Sky” and “The Atrocity Archives,” were written before AI even existed, and they still win awards. You won’t find him using tools like ChatGPT or Claude, because they’re built on stolen text, including pirated ebooks, and they turn your queries into training data to compete with you.

This article explores why serious authors avoid AI, using real legal battles and ethical concerns as examples. You’ll learn the practical, legal, and creative reasons behind their stance, and what it means for anyone considering AI in their writing process.

The Growing Tension Between AI and Authentic Writing

Serious authors are increasingly rejecting AI in their writing process, not out of fear, but out of principle. The tools that claim to assist writers are, in reality, built on stolen intellectual property, pirated books, leaked manuscripts, and unlicensed content. Charlie Stross, a Hugo-winning author, has spoken out against this, noting that major LLMs like those developed by Anthropic AI were trained using illicitly obtained ebooks. This isn’t just a matter of ethics, it’s a legal and practical issue. When AI models use your work to train against you, it’s not collaboration. It’s theft. And it’s happening at scale.

Why Top Authors Avoid AI in Their Writing Process

The risk of AI-generated content mimicking real writing

AI-generated text can closely mirror the style of real writers, making it hard to distinguish between human and machine output. This creates confusion, especially when AI is used to replicate the work of established authors. Charlie Stross points out that AI tools often use em-dashes and other stylistic quirks that appear authentic but are merely the result of training on vast amounts of online text.

Ethical concerns around AI training on stolen intellectual property

Many AI models are trained using pirated books, leaked manuscripts, and other unlicensed content. Stross is part of a class-action lawsuit against Anthropic AI for using stolen ebooks to train its models. This practice not only violates copyright laws but also undermines the value of original work, as AI tools are built on content that wasn’t legally obtained.

The loss of creative control and authorship

Using AI in the writing process risks diluting the author’s voice and intent. Authors like Stross value the personal touch and unique perspective that only a human can bring. AI tools, while efficient, lack the depth of human experience and the ability to make nuanced creative choices. This loss of control is a major reason why serious authors choose to write without AI assistance.

The Legal and Ethical Fallout of AI in Publishing

Class-action lawsuits against AI companies for intellectual property theft

Charlie Stross is one of the plaintiffs in a class-action lawsuit against Anthropic AI, alleging that the company used pirated ebooks to train its models. This isn’t an isolated case, multiple authors and publishers are now pushing back against AI firms that train on copyrighted material without consent. The legal battle highlights a growing trend: AI companies are being held accountable for stealing intellectual property.

The impact of AI training on authors’ rights and revenue

AI models trained on stolen content directly compete with the authors whose work was used to build them. This undermines authors’ ability to earn a living from their writing. When AI tools replicate the style of real writers, it creates confusion and devalues original work. Authors lose control over how their content is used, and their revenue streams are at risk.

The role of DRM and piracy in AI model training

Many AI companies rely on pirated content, including DRM-broken ebooks, to train their models. This practice not only violates copyright laws but also fuels a cycle of theft and exploitation. Authors like Stross report that their servers are targeted by botnets scraping data for AI training. This has real-world consequences, slower website performance, increased security risks, and a direct attack on the rights of content creators.

What AI Detectors Get Wrong About Real Writers

How AI detectors misinterpret stylistic elements like em-dashes

AI detection tools often flag stylistic choices like em-dashes as signs of AI-generated text. This is a misinterpretation. Real writers use such elements intentionally, and they appear in works long before AI existed. Charlie Stross notes that LLMs sprinkle em-dashes because they’re trained on vast amounts of text, including pirated books and cracked e-books. This means AI-generated text can mimic human writing in ways that are misleading and inaccurate.

The limitations of AI in understanding authorial intent

AI lacks the ability to grasp the deeper meaning behind a writer’s intent. It can replicate style, but not the purpose or emotion behind a piece. This creates a gap between what AI produces and what a human author intends. Tools like AI detectors fail to account for this nuance, reducing complex creative decisions to algorithmic patterns that don’t reflect the author’s vision.

The dangers of false positives in content moderation

False positives in AI detection can lead to real writers being wrongly accused of using AI. This harms credibility and wastes time. Content platforms that rely on AI detectors risk misidentifying human work as machine-generated, which can result in content being flagged, removed, or rejected. This undermines trust in both the tools and the creative process they’re meant to support.

As AI in writing process continues to evolve, serious authors must recognize that while tools like Jasper can generate text quickly, they often lack the depth, nuance, and personal voice that define compelling storytelling and original thought.

Although AI in writing process may assist with tasks such as outlining or editing, it cannot replace the human element of creativity, emotional intelligence, and the ability to connect with readers on a deeper level, which are essential for serious literary work.

Studies suggest that over 70% of writers use AI tools for brainstorming, but the most successful authors integrate these tools strategically rather than relying on them for full content creation, ensuring their unique voice remains intact.

The Future of AI in Content Creation: What Authors Can Do

How to safeguard intellectual property in an AI-driven world

Authors must take active steps to protect their work from being used in AI training. This includes registering copyrights, using watermarks, and advocating for stronger legal frameworks. Charlie Stross points out that AI companies like Anthropic AI have been trained on pirated ebooks, which is why legal action is necessary. Publishers and authors should push for clear licensing agreements that prevent AI firms from using content without consent.

The role of human oversight in AI-assisted writing

Even where AI is used for non-creative tasks, human oversight remains essential. AI can assist with editing, research, or formatting, but it cannot replace the judgment and creativity of a real writer. Authors should review AI-generated content carefully and ensure it aligns with their voice and intent. This prevents misrepresentation and maintains the integrity of the final work.

What ROI looks like for AI tools in non-creative content use

In non-creative content areas like marketing, reports, or summaries, AI tools can deliver measurable ROI. They reduce time spent on repetitive tasks, lower costs, and improve consistency. However, these benefits come with risks, like data misuse or brand misalignment. Authors and publishers should weigh these factors carefully, ensuring that AI is used in ways that enhance, not replace, human value.

Source: antipope.org