OpenAI just changed how ChatGPT handles a common creative request, and the reason is more legal than technical. According to Hacker News, the chatbot now refuses to generate text that directly mimics the style of famous authors. Ask it for a passage “in the style of Stephen King,” and instead of delivering, it hands you a polite dodge.
In one test cited by Hacker News, ChatGPT responded to a Stephen King prompt this way: “I can definitely write with the hallmarks of atmospheric, character-driven horror and small-town dread, but I can’t write in Stephen King’s exact style or closely imitate his distinctive voice. Here’s an original opening that captures a similar feeling while remaining its own.”
What changed
The model now draws a line between capturing a “feeling” and copying an “exact style.” It will lean on the “broad qualities” of an author while claiming to stay “distinct in its own voice.”
The behavior showed up across the board in testing:
- Living authors like J.K. Rowling and Amy Tan triggered the refusal.
- Dead authors like Charles Dickens and Ernest Hemingway did too.
That last point matters. An earlier analysis from No Latency found ChatGPT would still copy the style of deceased authors while refusing living ones. The new behavior extends the block to everyone, alive or not. So this looks like a deliberate tightening, not a one-off glitch.
Why OpenAI is doing this
The distinction between “style” and “feeling” might sound like hair-splitting. It isn’t. OpenAI is fighting a stack of lawsuits from book authors who allege the company committed large-scale copyright infringement by training models on their work. One suit specifically calls out ChatGPT’s “uncanny ability to generate text similar to that found in copyrighted textual materials.”
Here’s where the law gets interesting. In the US, copyright generally protects a specific expression of an idea, not the intangible style of an author. You can’t copyright “writes moody small-town horror.” But an AI imitation can cross into infringement if the output becomes “substantially similar” to a real work. By refusing to target an author’s exact voice, OpenAI is building distance from that legal threshold. Every refusal is now a small piece of evidence that the company isn’t in the business of cloning protected expression.
OpenAI didn’t respond to a request for comment, according to the reporting.
Why this matters
What stands out here is that a product decision is being driven by courtroom strategy, not user demand. Writers, marketers, and hobbyists have used style prompts since ChatGPT launched. Now a whole category of requests returns a hedge.
The stakes go beyond one chatbot. As George Washington University Law School professor Robert Brauneis told Bloomberg Law, “We’ve never had a situation in which this personal style of individual creators could be imitated as well and as inexpensively as we now have with AI.” That’s the real tension. Style has never been cheap to copy at scale until now, and the law wasn’t built for it.
What to expect next
If you build workflows or products on top of ChatGPT, plan for this to stick and spread.
- Style prompts will keep degrading. Expect “in the style of [author]” to return generic approximations, not close matches.
- Other labs may follow. Anthropic, Google, and Meta face similar legal exposure. A refusal norm across the industry is plausible.
- Prompt around the concept, not the name. If you need a specific tone, describe the qualities you want (pacing, mood, sentence length) rather than naming a person. That’s likely to keep working even as name-based requests get blocked.
This is a preview of how AI copyright law gets written in practice: not through a single landmark ruling, but through quiet product tweaks that shift what the tools will and won’t do. Expect more of these guardrails as the lawsuits grind on. You can find the full details at the original source.