This week, OpenAI introduced several new voice models to GPT-4o: Aiden, Ember, Zephyr, Orion, and Sky. However, the Sky voice model sparked controversy after actress Scarlett Johansson revealed that she was approached to license her voice for the new model but declined for “personal reasons.”

Johansson further reported that two days before the GPT-4o demonstration, Altman contacted her again, asking her to reconsider the proposal.

Before the actress could respond, the new feature was announced, followed by a post from Sam Altman on his X account (formerly Twitter) with the word “her.”

Understanding the Reference

In 2013, Scarlett Johansson voiced a bot in the movie “Her,” where a man falls in love with the generative model. Clearly, the Altman’s post was an allusion to the film and the actress’s voice, who had denied permission to use her voice.

The Controversy Grows

When the model was presented, what was anticipated by Altman’s post was confirmed. According to Johansson:

“Nine months later, my friends, family, and the general public all noticed how much the new ‘Sky’ system sounded like me. When I heard the released demonstration, I was shocked, angry, and disbelieving that Mr. Altman sought a voice that sounded so strangely like mine that my closest friends and media outlets couldn’t tell the difference.”

Scarlett then hired lawyers to contest the use of her voice by the Sky model. OpenAI denied that the model’s voice was based on the actress’s voice and apologized for the misunderstanding caused. The company stated:

“We believe AI voices should not deliberately mimic a celebrity’s distinctive voice — Sky’s voice is not an imitation of Scarlett Johansson and belongs to a different professional actress, using her own natural voice.”

Following this, the company announced that it would remove Sky’s voice from the list of available voices.

Timeline of Events

• First Contact: Altman contacts Scarlett to use her voice; Scarlett declines.
• Second Contact: Altman contacts her again months later.
• Launch: Before Scarlett could respond, the model is launched; Altman posts “her” on X.
• Reaction: One of the models’ voices is perceived as an unauthorized copy of the actress’s voice.
• Legal Action: Scarlett sues OpenAI.
• OpenAI’s Response: Sam Altman states that he did not intend to use the actress’s voice and believes AI voices should not deliberately mimic distinctive celebrity voices. (Here, we pretend he didn’t have that intention from the start, when he negotiated using the voice with the actress);
• Consequence: The voice is removed.

Reflections on Regulation and Ethics

Apparently, for the CEO of OpenAI, “no” doesn’t mean “no.” It means he will do what he wants anyway. Unfortunately, legal regulations are still very outdated and cannot keep up with the rapid technological evolution. Let’s not forget the hearing where a parliamentarian asked Mark Zuckerberg how Facebook makes money if registration on the platform is free.

There are no clear laws protecting our image in the virtual world. For now, this precaution has to be taken at an individual level. We are at a moment of risk of deliberate and unauthorized use of our own image. Scarlett Johansson shows complete awareness of this by stating:

“In an era where we are all fighting against deepfakes and for the protection of our own image, our own work, our own identities, I believe these are issues that deserve absolute clarity.”

What is your opinion on the matter? Leave your comment below and share on social media if you find this discussion relevant.

If you enjoy my work and want to support it, you can buy me a coffee!