OpenAI paused a ChatGPT voice after a likeness complaint
OpenAI's own account says it paused the Sky voice on 19 May 2024 after Scarlett Johansson objected, without confirming imitation.
- Historical event
- May 19, 2024
- First source published
- May 22, 2024
- Site publication
- September 18, 2026
What happened
OpenAI's own post, How the voices for ChatGPT were chosen, published 22 May 2024, states that the company 'paused the use of Sky in our products as of May 19, 2024.' The post gives a detailed timeline: OpenAI worked with casting directors from early 2023, received 'over 400 submissions from voice and screen actors,' and selected five voices, including Sky, through a five-month process. It also discloses that chief executive Sam Altman contacted actor Scarlett Johansson on 11 September 2023 to ask her to voice a sixth option, which she declined, and approached her again on 10 May 2024, shortly before a ChatGPT voice demonstration drew public comparisons between Sky and her voice.
What the documents show
The post states directly that 'the voice of Sky is not Scarlett Johansson's, and it was never intended to resemble hers,' attributing the resemblance the public heard to the actor cast rather than to any attempt to imitate a specific person. Altman is quoted apologising: 'We are sorry to Ms. Johansson that we didn't communicate better.' A separate, living document, OpenAI's voice mode help page, retrieved in September 2026, lists nine current standard voices — none named Sky — naming today's lineup as Arbor, Breeze, Cove, Ember, Juniper, Maple, Sol, Spruce and Vale, without stating what happened to the earlier five.
The mechanism
The mechanism here is casting and likeness, not text generation: a voice used in a companion-style product is a recorded human performance selected and licensed by the company, distinct from the language model choosing what to say. That distinction is why a pause is possible without retraining anything — the company can withdraw one recorded voice option while the underlying conversational model continues unchanged. It is also why a likeness complaint about a voice engages different questions, such as consent and casting practice, than a complaint about what a chatbot says.
What it leaves open
OpenAI's May 2024 post denies deliberate imitation but does not explain why public reaction so strongly read Sky as resembling Johansson despite that denial, and this record does not resolve that gap. The current help page's silence on Sky's status — neither confirming a permanent removal nor mentioning the 2024 episode — means this record can state that Sky is absent from the current list, but not, from these two documents alone, whether that is because of the May 2024 pause specifically or a later, separately motivated redesign of the voice lineup.
- Was a company's voice cast independently of any specific celebrity, and can that be verified beyond its own statement?
- Does 'paused' mean removed permanently, or temporarily withheld pending a decision?
- Does a current product page explain why an earlier, well-known feature is no longer listed?
The clearest fact this record can state is procedural: a company can withdraw a synthetic voice from a live product within days of a public complaint, independent of whether it accepts the underlying claim of resemblance. What that episode says about consent and likeness in AI voice products generally is a live question the two documents opened here do not settle.
Sources & reading trail
States the Sky voice was paused on 19 May 2024 and denies deliberate imitation of Scarlett Johansson.
Source published: 22 May 2024 · Retrieved: 16 September 2026
Living document: current voice roster (retrieved 2026-09-16) does not include Sky, without stating why.
Source published: Not established · Retrieved: 16 September 2026
Company documents, filings, studies and official records establish the record; the reading and the questions are Lovebot Journal editorial analysis. This retrospective draft does not imply the site published on the event date.
Continue reading
- GPT-4o's voice launch came with a warning about emotional reliance
- Hume's voice interface says it measures expression, not emotion
- Meta gave its assistant five actors' licensed voices, not clones
- Browse the complete the archive
Sources & reading trail
- How the voices for ChatGPT were chosen
Source published: May 22, 2024 · Retrieved: September 16, 2026 - Voice mode FAQ
Retrieved: September 16, 2026
The documents above establish the record. The reading and the questions are this publication’s editorial analysis, written after the fact.
Published September 18, 2026, not on the date of the event described.