AI Pulse by Inblix

ChatGPT now refuses to clone author voices as OpenAI braces for copyright rulings

Ars Technica AI · Jul 27, 2026 · 2 min read · Read original article →

Curated by the Inblix editorial team


Featured image for article: ChatGPT now refuses to clone author voices as OpenAI braces for copyright rulings

OpenAI has quietly flipped a switch. Ask ChatGPT to write in the exact style of Stephen King, J.K. Rowling, or Ernest Hemingway, and it now pumps the brakes. Instead of mimicking a specific voice, it offers a sanitized workaround—generating text that channels the “broad qualities” or “feeling” of those authors while insisting it can’t reproduce their distinctive style. Ars Technica confirmed the behavior this morning across multiple living and dead authors, a notable change from the platform’s previous willingness to churn out passable imitations on command.

The pivot isn’t about artistic integrity. It’s a legal shield. OpenAI is staring down a pile of lawsuits from authors who argue that training models on their copyrighted work amounts to mass infringement. One complaint specifically calls out the model’s “uncanny ability” to generate text similar to protected material. By forcing the AI to stay vague, OpenAI draws a line—however blurry—between emulating a genre and cloning an author. That distinction might sound like wordplay to a novelist, but in a courtroom, it’s everything. US copyright law protects specific expression, not general style, so stopping short of “substantial similarity” is the new imperative.

An analysis from No Latency earlier this month spotted the shift, noting that the refusal was initially limited to living authors. Now the net is cast wider, catching literary ghosts like Dickens alongside contemporary heavyweights. It’s a cautious move, and frankly, a clever one. Robert Brauneis, a law professor at George Washington University, told Bloomberg Law that we’ve never seen a tool that can duplicate personal style this cheaply and effectively. That novelty is exactly what makes the legal territory uncharted and terrifying for AI firms.

Still, the product feels diminished. The whole appeal of a language model is its ability to shapeshift. Telling it to be evocative but not too evocative is like hiring a forger who only paints in the “vibe” of Vermeer. For now, the compromises will keep coming. I suspect the line between inspiration and imitation will be litigated prompt by prompt, and the courts are nowhere near ready for that granularity.

💡 Key Takeaways

  1. OpenAI has restricted ChatGPT from imitating specific authorial voices, offering only broad stylistic features instead—a direct response to ongoing copyright lawsuits.
  2. The legal strategy hinges on a subtle distinction: US law protects specific expression, not general style, so preventing exact mimicry could limit liability for 'substantial similarity.'
  3. A law professor noted that cheap, high-fidelity AI style cloning is an entirely new legal problem, meaning the rules for what constitutes infringement here are still being written.
  4. This change applies to both living and dead authors, marking a significant product limitation that prioritizes legal defense over creative flexibility.

Keep reading: See related articles below for more coverage on this topic.

Get smarter about AI

The sharpest AI news, curated daily. Delivered free to your inbox.

Learn more

Glossary terms

← Back to all articles