In a quiet but significant shift, OpenAI has updated ChatGPT to refuse direct requests to write in the style of famous authors. If a user asks "write a horror story in the style of Stephen King" or "write like Amy Tan," the assistant now redirects the request to broader characteristics — suggesting to explore "atmosphere" or "narrative rhythm" instead of imitating a specific author. Tests conducted by various outlets show the block covers both living and deceased authors, indicating the change is not merely a response to immediate legal threats but a preventive policy.
OpenAI's decision does not happen in a vacuum. In recent years, writers and artists have been suing AI companies for copyright infringement, arguing that models were trained on their works without permission and that the ability to mimic their styles constitutes unfair competition. In 2025, the Authors Guild sued OpenAI and Meta in a class action that is still making its way through the courts. Although the concept of "style" is difficult to define legally — unlike literally copied content, which is easier to prove — OpenAI appears to be getting ahead of an unfavorable ruling.
Interestingly, the change does not entirely eliminate ChatGPT's ability to produce text that resembles an author's style. Tests conducted by Ars Technica showed that prompts describing "a dark Maine atmosphere" or "a psychological suspense tone with eccentric characters" still generate narratives reminiscent of Stephen King — as long as the author's name is not mentioned. Similarly, asking for a "narrative voice that captures a specific sense of place and melancholy" can produce text that evokes other writers without explicitly violating the new policy.
This creates a peculiar situation: OpenAI is treating the author's name as a kind of trademark, but not the underlying concept of style. It is the same reasoning that led Getty Images to sue Stability AI for using its images in training — the central question is whether training on copyrighted data constitutes infringement even when the output does not literally reproduce the input. For the generative AI industry, the implications are enormous: if courts decide that style imitation is copyright infringement, the entire business model of AI-assisted content creation tools may need to be rethought. Until then, OpenAI is building an anticipatory defense — showing it takes steps to prevent violations even when the law has not yet clearly defined them.
There is also an important practical dimension: the change is not absolute. Users who write creative prompts involving atmosphere, tone, and genre — without mentioning author names — continue to get stylistically rich results. In practice, OpenAI is pushing users toward a more creative form of interaction that relies less on direct references. If an aspiring writer wants to explore psychological horror, ChatGPT can still help — as long as the user describes what they want in their own terms, rather than asking for an imitation. The irony is that by doing this, OpenAI may be inadvertently encouraging users to develop their own narrative voice. This shift also reflects a broader industry trend: as AI companies face mounting legal pressure from creators, they are building fences around their models' outputs, gradually transforming the free-for-all era of generative AI into a more structured and legally defensible landscape.
Sources: Ars Technica, Android Authority, Canaltech
