OpenAI Voice Mode Resembling Scarlett Johansson Without Consent (2024)

Incident Summary

In May 2024, OpenAI publicly demonstrated GPT-4o, its multimodal AI model, featuring a voice interaction mode with several voice options including one called “Sky.”^[1] Actress Scarlett Johansson subsequently issued a public statement asserting that the Sky voice bore a striking resemblance to her own voice, and that she had twice declined OpenAI CEO Sam Altman’s requests to license her voice for ChatGPT — once in September 2023 and again just two days before the GPT-4o launch.^[1][3]

Johansson stated she was “shocked, angered, and in disbelief” upon hearing the Sky voice, noting that friends and media outlets could not distinguish it from her own voice.^[1] She also drew attention to Altman’s social media post of the single word “her” following the GPT-4o demonstration — a reference to the 2013 film Her, in which Johansson voiced an AI operating system — which she interpreted as an implicit acknowledgment of the resemblance.^[3][4]

OpenAI responded by pausing the Sky voice and publishing a blog post stating that the voice was performed by a different, unnamed professional actress and was not intended to resemble Johansson.^[2] Johansson retained legal counsel, and the matter prompted broader discussion about AI voice synthesis, consent, and the adequacy of existing right-of-publicity and intellectual property protections in the context of AI-generated voice likenesses.

Key Facts

• Parties involved: Scarlett Johansson (claimant); OpenAI (respondent)
• Voice in question: “Sky,” one of several voices available in ChatGPT’s voice mode
• Declined requests: Johansson declined Altman’s requests to license her voice on two occasions (September 2023 and May 2024)
• Social media signal: Altman posted the word “her” following the GPT-4o demo, referencing the film in which Johansson voiced an AI
• OpenAI’s position: Sky was voiced by a different professional actress and was not an imitation of Johansson
• Outcome: Sky voice paused; legal engagement initiated

Threat Patterns Involved

Primary: Biometric Exploitation — The incident raised the question of whether AI systems can exploit an individual’s vocal biometric identity by producing synthetic voices that closely resemble a specific person’s voice, particularly when that person has explicitly declined consent.

Secondary: Deepfake Identity Hijacking — Regardless of intent, the production of an AI voice closely resembling a recognizable public figure’s voice creates a form of identity appropriation, as the public may reasonably associate the AI voice with that individual.

Significance

1. Consent and voice likeness in the AI era. The incident crystallized a legal and ethical question at the intersection of AI and identity rights: whether existing intellectual property, right-of-publicity, and consent frameworks are adequate to address AI systems that can replicate or closely approximate an individual’s voice.
2. High-profile test case for voice cloning norms. As one of the most publicly visible disputes over AI voice synthesis, the Johansson-OpenAI incident has been widely cited in legal, regulatory, and industry discussions about establishing norms for AI-generated voice likenesses.
3. Corporate communication and implied association. Altman’s “her” social media post, interpreted by many observers as a deliberate reference to Johansson’s role in the 2013 film, raised questions about corporate responsibility when public communications create or reinforce associations between AI products and specific individuals.
4. Impetus for voice protection legislation. The incident contributed to legislative interest in voice likeness protections, including proposed state and federal laws in the United States addressing unauthorized AI replication of individuals’ voices.