Updated
Updated · Ars Technica · Jul 27
ChatGPT Blocks Author Style Mimicry as OpenAI Fights Copyright Lawsuits
Updated
Updated · Ars Technica · Jul 27

ChatGPT Blocks Author Style Mimicry as OpenAI Fights Copyright Lawsuits

2 articles · Updated · Ars Technica · Jul 27

Summary

  • ChatGPT now refuses prompts to write in the exact style of named authors, instead offering text that borrows broad qualities while keeping a distinct voice.
  • Tests cited by Ars Technica showed the shift across living and dead writers, including Stephen King, J.K. Rowling, Amy Tan, Charles Dickens and Ernest Hemingway.
  • The change could matter legally because authors suing OpenAI have pointed to ChatGPT’s ability to generate text resembling copyrighted works trained into its models.
  • U.S. copyright law generally protects specific expression rather than style alone, but close AI imitation can still become infringing if it is substantially similar to original works.

Insights

If human writers can legally mimic an author's style, why is ChatGPT suddenly banned from doing the exact same thing?
Are OpenAI's new guardrails genuinely protecting authors, or just a clever legal shield to hide their training data secrets?