Scarlett Johansson vs OpenAI: What Really Happened with the ChatGPT Sky Voice Controversy
In May 2024, one of the most high-profile clashes between Hollywood and artificial intelligence erupted when Scarlett Johansson accused OpenAI of using a ChatGPT voice that sounded uncannily like her own. The dispute centered on a voice option called “Sky,” which many listeners immediately associated with Johansson’s performance as the AI assistant Samantha in the 2013 film Her. What followed was a rapid sequence of legal letters, a public statement from the actress, and OpenAI’s decision to pull the voice. More than two years later, the episode still stands as a defining early example of the unresolved tensions around AI voice cloning, celebrity consent, and the limits of existing law.
The cultural backdrop made the controversy almost inevitable. Spike Jonze’s Her remains one of the most influential cinematic portraits of human-AI relationships. Johansson’s warm, slightly raspy, emotionally intimate voice made the fictional operating system feel alive and desirable. OpenAI CEO Sam Altman has repeatedly described the film as a favorite. When the company began developing more natural, conversational voices for ChatGPT, the cultural reference point was obvious to almost everyone watching.
According to Johansson’s account, Altman first approached her in September 2023 with an offer to voice the ChatGPT system. He reportedly framed the invitation as an opportunity to bridge the gap between technology companies and creatives, and to help the public feel more comfortable with the coming shift toward AI. After careful consideration, Johansson declined for personal reasons. Nine months later, when OpenAI demonstrated its new GPT-4o model and accompanying voice mode on May 13, 2024, one of the five available voices — Sky — struck many people, including Johansson’s own friends and family, as remarkably similar to her.
The timing of Altman’s second outreach intensified the sense of betrayal. Two days before the public demo, he contacted Johansson’s agent and asked her to reconsider. Before the two sides could connect, the demonstration went live. During the rollout, Altman posted a single-word message on X: “her.” To Johansson and many observers, the tweet appeared to acknowledge the deliberate parallel.
Johansson responded with a detailed public statement that mixed personal anger with broader concerns. She said she was “shocked, angered and in disbelief” that Altman would pursue a voice so similar to hers that even close friends and news outlets could not tell the difference. She noted that Altman’s “her” post insinuated the resemblance was intentional. After the demo, she hired legal counsel, who sent two letters to Altman and OpenAI demanding a precise explanation of how the Sky voice had been created. Only after those letters, she said, did the company “reluctantly” agree to take the voice down. Johansson closed by calling for greater transparency and “appropriate legislation” to protect individual rights in an era of deepfakes and AI-generated likenesses.
OpenAI pushed back firmly on the claim of deliberate imitation. In statements issued after Johansson went public, the company and Altman insisted that Sky was never intended to resemble her. They said the voice belonged to a different professional actress using her own natural speaking voice, and that this actress had been cast before any outreach to Johansson. OpenAI had developed five voices through a months-long process that began in 2023. The company maintained that AI voices should not deliberately mimic a celebrity’s distinctive voice and that it had paused Sky out of respect for Johansson while apologizing for poor communication.
Independent reporting largely supported OpenAI’s timeline on the casting. The Washington Post reviewed documents, recordings, and spoke with people involved in the process, including the actress’s agent (who spoke on condition of anonymity). The reporting confirmed that a separate actress had been hired months earlier, that neither Johansson nor the film Her had been mentioned during casting, and that the actress’s natural voice closely matched the final Sky product. The name “Sky” itself was chosen to evoke a cool, airy quality rather than any specific celebrity association.
Despite the competing narratives, the practical outcome was swift. OpenAI paused the Sky voice within days of receiving the legal letters. No lawsuit was ever filed. Because the company withdrew the feature before litigation began, the courts never ruled on the core questions the dispute raised. As of August 2026, those questions remain unanswered. There is still no comprehensive federal law in the United States that clearly governs AI-generated voice sound-alikes. Right-of-publicity protections exist primarily at the state level. California’s statute, for example, covers unauthorized commercial use of a person’s name, voice, signature, photograph, or likeness, but courts have not yet fully tested how those rules apply when an AI system approximates a celebrity’s voice without directly copying their recordings.
The legal uncertainty is significant. Historical cases such as Bette Midler’s successful lawsuit against Ford in the 1980s established that a distinctive voice can be protected against deliberate imitation for commercial purposes. Whether those principles extend cleanly to modern generative AI remains an open question. Johansson’s lawyers never forced a definitive test, leaving the industry without clear precedent.
Beyond the specific facts of the Sky voice, the controversy crystallized larger anxieties that continue to shape debates about AI. Performers worry that their voices and likenesses can be approximated without consent or compensation, potentially diluting the value of their work. Technology companies argue that restricting sound-alike voices too tightly could stifle innovation and that coincidental similarities are inevitable when training models on large datasets of human speech. The public, meanwhile, is left navigating a growing volume of synthetic media with limited tools to verify authenticity or enforce boundaries.
The episode also highlighted problems of communication and optics. Even if OpenAI’s casting process was independent and the resemblance unintentional, the combination of Altman’s prior outreach, the timing of the second request, and the “her” tweet created a powerful perception of bad faith. In an environment already marked by distrust between creative industries and AI labs, perception often matters as much as technical reality.
Two years on, the Sky voice remains offline. OpenAI has continued to expand its voice offerings with other options, while the broader industry has moved forward with increasingly sophisticated voice synthesis. Legislation in various jurisdictions has begun to address deepfakes and AI likeness rights, but comprehensive federal rules specifically covering voice cloning still lag. Johansson’s public stand helped elevate the issue, yet the underlying legal and ethical questions she raised have not been settled.
The Scarlett Johansson–OpenAI dispute was never simply about one chatbot voice. It was an early, highly visible collision between the speed of AI product development and the slower machinery of consent, law, and cultural expectation. The company removed the contested feature under pressure. The actress made her objections clear. The rest of the industry is still working out what the rules should be.
About The Author
