Actress Scarlett Johansson has recently come forward to express her shock and anger after discovering that OpenAI had allegedly recreated her voice without her consent for their new ChatGPT system. Johansson personally criticized the company’s CEO, Sam Altman, for insinuating that she was the voice named ‘Sky’ by posting the word “her” on X, referencing a film in which she voiced an AI that a human fell in love with.
According to Johansson, Altman had approached her in September 2023 to voice the ChatGPT 4.0 system, but she declined due to personal reasons. In her statement, she shared that Altman had suggested her voice could bridge the gap between tech companies and creatives and help consumers feel comfortable with the growing presence of AI. However, she declined the offer.
Nine months later, Johansson was made aware of how much the released demo of OpenAI’s ChatGPT 4.0 sounded like her, causing her to feel shocked, angry, and in disbelief. She stated that even her closest friends and news outlets could not tell the difference between her voice and the one used by OpenAI’s Sky.
On May 13, Altman posted the word “her” on X, which Johansson believed was a deliberate reference to the 2013 film “Her” in which she voiced an AI that forms an intimate relationship with a human. The post has since been seen by over 10.2 million people and remains on Altman’s X feed.
Johansson also revealed that two days before the ChatGPT 4.0 demo was released, Altman had contacted her agent, asking her to reconsider voicing the system. However, before they could connect, the system had already been released.
In response to Johansson’s statements, OpenAI has paused the use of the controversial voice “out of respect for Ms. Johansson,” according to Altman’s statement on Tuesday. The company offers five voices for its ChatGPT service and has denied that Sky was based on Johansson’s voice, stating that the voice actor behind Sky was cast before any outreach to the actress.
OpenAI released a blog post detailing their process for choosing voice actors, stating that they received over 400 submissions from screen and voice actors in 2023. They looked for voices that were diverse, timeless, approachable, warm, engaging, and had a rich, confident, and charismatic tone. The selected actors were then flown to San Francisco for recording sessions and launched into ChatGPT in September 2023.
The AI company also addressed the issue of deepfakes, stating that they find them unacceptable and are working to prevent their synthetic voices from copying those of real people. Deepfakes are digital imitations of people made to seem realistic using AI.
The US screen actor’s guild, SAG-AFTRA, has shown support for Johansson’s claims, stating that their members are among the most talented and recognizable individuals and that they are strongly advocating for federal legislation to protect their voices and likenesses from unauthorized digital replication.
OpenAI has faced legal trouble before, with The New York Times suing the company at the end of last year for unlawfully using their articles to train and create ChatGPT. The suit alleges that the AI text model now competes with the newspaper as a source of reliable information and poses a threat to their ability to provide such a service.
