Detect Deepfakesby Resemble AI
Deepfake case study · Audio

Johansson Turned Down OpenAI Twice: Still No Federal Law…

Scarlett Johansson challenged OpenAI after the company released a ChatGPT voice that closely resembled her own despite her declining two requests to…

Incident date
May 2024
Target
Scarlett Johansson
Updated Aug 19, 2026 · 1 min read

In May 2024, OpenAI released a voice update for its GPT-4o model featuring an AI voice named Sky. Following the release, actress Scarlett Johansson issued a public statement expressing shock and anger, noting that the voice was so similar to her own that her closest friends and news outlets could not distinguish between them.

What happened

OpenAI CEO Sam Altman had approached Johansson in September 2023 to voice the ChatGPT system, a request she declined. Two days before the May 2024 demonstration, Altman contacted her agent again, but she declined once more. Despite these refusals, the demonstration proceeded with the Sky voice. Following legal intervention from Johansson’s representatives, OpenAI withdrew the voice.

OpenAI maintained that the voice belonged to a different professional actress hired before they contacted Johansson, and the actress's agent confirmed that Johansson was not mentioned during the audition process. However, research from Arizona State University found that the Sky voice profile matched Johansson’s more closely than 98% of other actresses tested.

This incident highlights a significant legal tension regarding the right of publicity. While cases like Midler v. Ford Motor Co. and Waits v. Frito-Lay established precedents against "sound-alike" commercial uses, there is no federal right of publicity in the United States. Protections remain at the state level, such as California Civil Code § 3344 or Tennessee’s ELVIS Act. The Johansson case remains complex because it involves AI systems generating a voice that sounds like a specific individual without necessarily copying existing recordings, a scenario existing legal doctrine was not originally designed to address. As of August 2026, no court had issued a ruling on Johansson's specific claim, leaving a structural gap in how vocal labor and likeness are protected in the age of generative AI.

Sources