Scarlett Johansson’s Claims About OpenAI’s ‘Sky’ Voice

Quick answer

OpenAI introduced a new AI voice called 'Sky' for its ChatGPT assistant in May 2024. Actress Scarlett Johansson claimed the voice bore an 'eerie' resemblance to her own. She stated she had previously declined an offer from OpenAI's CEO to license her voice. OpenAI paused the 'Sky' voice following her public statement.

What is OpenAI's 'Sky' voice?

In May 2024, OpenAI unveiled new voice and vision capabilities for ChatGPT. This update allowed users to interact with the AI assistant using spoken commands and receive responses in a natural-sounding voice. OpenAI introduced five distinct voices as part of this new interaction mode.

The 'Sky' voice was one of these options. It was designed to offer an expressive, human-like tone, aiming to make conversations with ChatGPT feel more intuitive and less robotic. These voices were made available to all ChatGPT users, including those on the free tier and Plus subscribers. The goal was to provide a more engaging and accessible experience for interacting with the large language model.

Scarlett Johansson's claims about the voice

Following OpenAI's demonstration of the new voice capabilities in May 2024, actress Scarlett Johansson issued a public statement on 20 May 2024. In her statement, she asserted that the 'Sky' voice sounded remarkably similar to her own. She described the resemblance as 'eerie'.

Johansson also revealed a prior engagement with OpenAI. She stated that Sam Altman, OpenAI's CEO, had approached her several months earlier with an offer. Altman reportedly asked her to license her voice for use in OpenAI's AI systems. Johansson declined this offer.

Her statement conveyed concern that OpenAI had proceeded to create a synthetic voice highly similar to hers, despite her refusal to collaborate. She noted that the resemblance caused confusion among her close friends and several news outlets. These sources also observed the similarity between the 'Sky' voice and Johansson's distinctive vocal qualities. The actress implied a deliberate attempt to replicate her voice without her consent.

OpenAI's response and subsequent actions

OpenAI initially responded to the growing public discussion by stating that the 'Sky' voice belonged to a professional voice actress. The company maintained that the voice was not an imitation of Scarlett Johansson's or any other celebrity's voice. They also said they do not create AI voices that intentionally mimic celebrity voices.

However, following Johansson's public statement and the widespread media attention, OpenAI announced a change in its approach. On 20 May 2024, the company confirmed it would pause the use of the 'Sky' voice. This decision was made to address the concerns raised and out of respect for Johansson.

Sam Altman, OpenAI's CEO, also made a comment that added to the perception of intentional similarity. Prior to the controversy, he had tweeted a single word, 'her', which many interpreted as a reference to the 2013 film 'Her'. In that movie, Scarlett Johansson voiced an AI assistant. This tweet was later removed or deleted. OpenAI has stated it will be more transparent in the future regarding how it selects and develops its AI voices.

What this changes for AI users and developers

For people who use AI assistants, this incident highlights important questions about the origin and ethical sourcing of synthetic voices. When an AI voice sounds very natural or familiar, users might now pause to consider if it is based on someone's actual voice without their permission. It also underscores the potential for AI technology to create confusion regarding identity and intellectual property.

For companies developing AI products, this event sets a precedent. It puts greater pressure on them to ensure clear consent and transparency in their voice creation processes. Developers may need to demonstrate that their AI voices are distinct and do not infringe on individual voice rights. This could involve more rigorous verification of voice actors and ensuring that synthetic voices are not designed to imitate specific, recognisable individuals without explicit licensing agreements.

The controversy signals a shift in focus. The industry may now move towards stricter guidelines or even regulations on how AI voices are developed and deployed. This aims to protect the voices and identities of individuals in an era where synthetic media is increasingly sophisticated. It could mean more due diligence for developers and more assurance for consumers about the origins of the voices they hear.