ChatGPT stops imitating authors: OpenAI protects itself from copyright

What happened?

OpenAI has implemented a change in ChatGPT that prevents the chatbot from generating text 'in the style of' a specific author. For example, if asked to 'write a chapter in the style of Stephen King,' the model responds that it cannot do so, offering instead a text with 'the characteristics of atmospheric horror and small-town suspense,' but insisting that it is an original creation.

The change was detected by Ars Technica and confirmed by The Next Web. Although OpenAI has not issued an official statement, it is speculated that the measure aims to prevent copyright infringement lawsuits, in a context where writers like John Grisham or George R.R. Martin have sued the company for training its models with protected works.

Why is it important?

This decision marks a turning point in the relationship between generative AI and copyright. Until now, ChatGPT was known for its ability to emulate literary styles, which made it popular among writers and enthusiasts. However, this use bordered on the legal limit: if the model reproduces patterns of an author without a license, it could be considered an unauthorized derivative work.

For users, the loss of this functionality reduces the assistant's creative appeal. They will no longer be able to ask for 'a poem like Pablo Neruda' or 'a horror story like Lovecraft.' Instead, ChatGPT offers generic descriptions of styles but avoids direct imitation.

Consequences for the market

• For creators: The measure does not satisfy the suing authors, who demand compensation for the use of their works in training. The change only affects the output, not the training data.
• For competitors: Other AIs like Claude or Gemini could follow suit to avoid litigation. However, some smaller startups could exploit this gap by offering stylistic imitation as a differentiator.
• For users: The versatility of ChatGPT in creative tasks is reduced. Writers who used the AI as an inspiration tool will need to seek alternatives or adjust their prompts.

What should readers know?

The change is global and affects all versions of ChatGPT, including the API. There is no way to bypass it with prompt engineering; the model explicitly rejects requests to imitate protected authors. However, it is still possible to ask for 'a text with a style similar to...' without mentioning specific names, although the results are less precise.

OpenAI is in a delicate position: on one hand, it must protect its business from lawsuits; on the other, it does not want to alienate creative users. It is likely that in the future we will see a licensing system that allows authorized imitation, similar to the agreements it already has with some media outlets.

"OpenAI has opted for caution: it prefers to limit functionality now rather than risk an adverse ruling that could cost billions. But the question remains: can AI be creative without imitating?"

Legal context and precedents

The class-action lawsuit by authors against OpenAI, filed in 2023, alleges that the company used protected books without permission to train its models. Although the change in ChatGPT does not resolve the substance of the matter, it shows that the company is willing to make concessions on the product to improve its negotiating position.

In parallel, the European Union is advancing in AI regulation, which will require transparency in training data. This type of adjustment could become a regulatory requirement, not just a business decision.

Conclusion

OpenAI's decision to limit author imitation in ChatGPT is a defensive move amid the copyright legal storm. While it protects the company, it reduces the assistant's creative value for users. The future of generative AI will depend on how this conflict is resolved: through licenses, regulation, or new technologies that allow originality without infringing rights.