OpenAI pulls ChatGPT's Sky voice after users said it sounded like Scarlett Johansson, saying it is an actress' voice and wasn't chosen to be an “imitation”
OpenAI is working to pause the use of the Sky voice from an audible version of ChatGPT after users said that it sounded too much like actress Scarlett Johansson.
Context & Ripple Effects
The pause landed as OpenAI was showcasing a more conversational voice assistant, built around end-to-end speech interactions that handle interruptions and emotional cues. That makes voice identity part of the product experience rather than a cosmetic interface choice.
The dispute quickly expanded beyond listener comparison: Johansson said she had declined an offer, while later reporting described OpenAI's separate casting of the Sky voice actress. The episode put the company’s selection process and public presentation under unusual scrutiny.
First-order effects
- OpenAI removes or pauses Sky from ChatGPT’s audible experience, limiting a prominent voice option while it addresses the resemblance claims.
- Johansson’s objection turns a product-design choice into a reputational issue for OpenAI, despite the company’s statement that Sky was not selected as an imitation.
Second-order effects
- Voice-AI developers have a clearer incentive to vet not only consent and contracts but also whether a deployed voice is recognizably associated with a public figure.
- The rollout of more natural, interruption-capable voice assistants faces heightened scrutiny because humanlike delivery makes perceived likeness more salient to users.
Third-order effects
- As conversational AI becomes more lifelike, voice provenance and performer consent are likely to become core product-governance requirements, alongside model safety reviews.
- The case points toward a market in which distinctive synthetic voices may need clearer disclosure, documentation, and licensing boundaries; how far those practices become formalized remains unsettled.
The trend: Humanlike voice interfaces are shifting AI competition toward questions of identity, consent, and provenance as much as conversational capability.