What Happened
On May 20, 2024, actress Scarlett Johansson released a public statement revealing that OpenAI CEO Sam Altman had personally contacted her in September 2023 asking her to voice the ChatGPT system, which she declined. When OpenAI launched its GPT-4o demo on May 13, 2024, one of the voices offered, called “Sky,” sounded strikingly similar to Johansson’s voice. Altman had even posted a single word — “her” — on X during the demo, referencing the 2013 film in which Johansson voiced an AI assistant.
Johansson stated she was “shocked, angered, and in disbelief” that OpenAI had apparently created a voice that sounded so similar to hers that her closest friends and family could not tell the difference. She retained legal counsel and demanded transparency about how the voice was created.
Timeline
Altman contacted Johansson in September 2023; she declined. OpenAI launched the GPT-4o demo with Sky voice on May 13, 2024. Johansson heard the demo and contacted OpenAI through legal counsel. Altman texted Johansson asking her to reconsider just two days before the launch. On May 20, Johansson released her public statement. OpenAI paused the Sky voice the same day and stated it was voiced by a different professional actress.
Impact
The controversy damaged OpenAI’s reputation significantly, raising questions about the company’s ethics and respect for consent. The “her” social media post made it difficult for OpenAI to credibly claim the similarity was unintentional. The incident raised broader questions about voice rights, likeness rights, and whether AI companies can legally create voices that closely mimic specific individuals who have refused consent.
The episode also highlighted the power imbalance between AI companies and individuals whose likenesses or voices might be replicated without permission.
Response
OpenAI stated that Sky was voiced by a different professional actress (whose identity they initially kept private to protect her) and that the similarity to Johansson was coincidental. The company paused the voice out of respect for Johansson’s concerns but maintained it had not violated any laws. CTO Mira Murati stated the voice was cast independently.
Lessons Learned
The Johansson incident demonstrated that consent and transparency are essential when AI companies create products that may evoke specific individuals. Even if the legal claim is defensible (the voice was technically performed by someone else), the perception of mimicry after explicit refusal created an enormous reputational and ethical problem. The case established that approaching a celebrity for their voice, being refused, and then releasing a similar-sounding alternative creates an inference of intent that is difficult to overcome.
The incident accelerated discussions about voice rights legislation and the need for clearer legal frameworks governing AI voice synthesis.