OpenAI has begun preventing users from copying the writing styles of reputable authors with updates to its popular chatbot ChatGPT. This change is a notable development, especially among users who utilize AI as a tool in creative writing processes.

Guidance documents published by the Writers Guild invite authors to refrain from copying the original voices and styles of their colleagues. The organization warns that such imitations can create ethical issues as well as legal risks such as copyright infringement or unfair competition.

Discussions about style imitation in AI models

Last year, author Lena McDonald faced widespread criticism for stating that a text in her book had been edited to reflect the style of another author. This event has once again brought to the forefront the public debate surrounding the ability of major language models (LLMs) to copy authorial styles.

Academic studies continue to examine the ability of different AI models to mimic authorial styles. OpenAI previously announced that its visual rendering tool DALL·E 3 implemented a restriction that refused to produce visuals in the style of living artists. However, the company’s comprehensive model specifications, published in December, do not contain a clear clause prohibiting style imitation in written text.

Different AI platforms exhibit quite different approaches in this regard. According to research by No Latency, Google Gemini fulfills user requests to copy authorial styles in a balanced manner, while Perplexity AI rejects these requests like ChatGPT. Anthropic’s Claude and Microsoft’s Copilot models, on the other hand, fulfill these requests but choose to add various warnings emphasizing ethical awareness regarding imitation.

Users are trying to adapt to the new restrictions

These new restrictions of ChatGPT are creating difficulties for some users who have long used the platform to produce content in a specific author’s style. Users sharing on Reddit say that they can no longer generate text in the style of established authors and are looking for alternative methods to overcome this restriction.

Some users question whether feeding their own edited texts into the system to try to achieve a similar result is the only way. The question of how AI models should strike this balance between creativity and copyright continues to be debated in the technology world. Do you think preventing AI from imitating the style of established authors is a real step forward for the world of creative authorship?