ChatGPT starts blocking direct requests to copy an author's style
New behavior capturing a writer's "broad qualities" could have legal implications.
ChatGPT has begun refusing prompts that ask it to clone a named author's voice. The refusal is polite, even literary: it will capture the "broad qualities" of a Stephen King or a J.K. Rowling, but it will not imitate the distinctive style. The shift is small in surface and large in implication.
The interesting detail is the asymmetry. Ars Technica's testing, cross-checked against an earlier No Latency analysis, shows ChatGPT dodging requests for living authors while still complying for the deceased. That is not a technical limitation. It is a legal calculation dressed up as a safety feature.
Copyright protects expression, not ideas. A living author's voice, cadence, and stylistic fingerprints can be argued as protectable trade dress for their professional identity. A dead author's estate holds the rights to the published works, but the persona itself is harder to police. OpenAI has drawn the line exactly where the litigation risk is highest and the enforcement machinery is most active.
For the labor market, the more relevant question is what this means for the millions of writers, marketers, and content operators who have been quietly using ChatGPT as a ghostwriter with a famous name attached. The tool is being recalibrated to deny the most legally exposed use case while preserving the general-purpose productivity pitch. The capability is not being removed; the liability is being rerouted.
Watch the next move. If competitors keep cloning voices freely, OpenAI's restraint becomes a marketing liability. If they follow suit, the entire category of "write like X" prompting quietly disappears from the consumer tier and migrates into enterprise contracts with indemnification clauses. Either outcome reshapes who gets to monetize style at scale.