Scarlett Johansson Alleges OpenAI's Sky Imitates Her Voice Without Licensing

OpenAI unveiled a voice assistant with a voice resembling Scarlett Johansson's, despite her refusal to license her voice. Johansson claimed the assistant, "Sky," sounded "eerily similar" to her voice, leading her to seek legal action. OpenAI suspended Sky, asserting the voice was from a different actress.

Risk classification

• Primary risk domain: 6 Socioeconomic & Environmental
• Primary risk subdomain: 6.3 Economic and cultural devaluation of human effort

The incident involves the alleged unauthorized imitation of a celebrity's distinctive voice and likeness, highlighting anxieties about the devaluation of human creative work and the lack of consent and compensation for artists.

Additional risk subdomains

• 6.5 Governance failure: The incident highlighted the lack of adequate federal regulatory frameworks and legal protections for individuals' voices and likenesses against AI replication.
• 5.1 Overreliance and unsafe use: The flirty, highly human-like voice design raised concerns about users anthropomorphizing the AI and forming unhealthy emotional attachments.

Causal factors

• Entity: Human
• Intent: Unintentional
• Timing: Post-deployment

The incident arose from human decisions during the casting, development, and marketing of the ChatGPT voice feature, resulting in an unexpected public backlash and legal dispute after its public demonstration.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Limited Risk: The system is a chatbot and generates synthetic voice outputs. Under the EU AI Act, chatbots and AI systems generating synthetic content (like deepfakes or synthetic voices) are classified as Limited Risk, requiring transparency so users know they are interacting with an AI.

AI system and alleged parties

• AI system: ChatGPT 4o (OpenAI)
• AI purpose: AI Voice Assistant; Voice Generation
• Behaviour type: Assistant
• Alleged developer: Sam Altman, OpenAI
• Alleged deployer: Sky voice assistant, Sam Altman, OpenAI
• Alleged harmed parties: Scarlett Johansson

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Negligible. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Negligible
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Negligible, indirect Negligible
• Differential treatment: direct Negligible, indirect Negligible
• Civil rights: direct Negligible, indirect Negligible
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Negligible, indirect Negligible
• Epistemic: direct Negligible, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

People affected

• Occurrences reported: 1
• People reportedly harmed: 1
• People reportedly exposed: 1

Potential causes

Management

• Executive Hubris and Outreach: Altman pursued Johansson twice, creating a perception of deliberate copying.
• Poor Internal Coordination: Few employees knew of Altman's outreach, leading to misaligned actions.
• Provocative Public Marketing: Altman's 'her' tweet publicly linked the product to Johansson's character.

Technology

• Highly Realistic Voice Generation: GPT-4o voice technology allows highly realistic, humanlike vocal expression.
• Flirtatious Conversational Tone: The model was designed to use a playful, expressive tone that drew scrutiny.

Data Inputs

• Similar Natural Voice Casting: Hired actress had a natural voice with a similar breathiness and huskiness.
• Lack of Explicit Voice Data Consent: Industry standard of training models on internet data without explicit consent.

Human Factors

• Public Association with Her Movie: Users strongly associated the warm, breathy voice with the movie Her.
• Perception of Male Fantasy Voice: Critics felt the coquettish tone reflected a male developer's fantasy.
• Fear of Likeness Appropriation: High sensitivity among creatives regarding AI replicating their work.

Process and Methods

• Lack of External Verification: No independent process to verify that the voice did not mimic a celebrity.
• Inadequate Communication Process: Failed to communicate the casting process transparently to Johansson's team.
• Rapid Release Cycle: Released the demo before establishing clear legal and ethical boundaries.

Regulatory Environment

• Scant Voice Protection Regulations: Federal laws have not matured to protect a person's voice from AI replication.
• Wild West Legal Landscape: Absence of clear, nationwide rules governing AI training on human likeness.
• Non-binding Local Protections: Voice protections vary widely by state, creating legal loopholes.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide highly detailed, consistent accounts of the timeline, the statements from both Scarlett Johansson and OpenAI, the legal context, and the technical details of the GPT-4o voice feature. There is minimal ambiguity regarding the core events, although OpenAI's internal intent remains a point of public dispute.
• Ambiguities identified: Whether OpenAI internally intended to copy Johansson's voice or if the similarity was purely coincidental due to casting a natural soundalike.
• Alternative interpretations: None that significantly alter the classification of the incident as a likeness/publicity rights dispute.

OpenAI's release and subsequent suspension of the 'Sky' voice assistant, which closely resembled actress Scarlett Johansson, sparked significant intellectual property and ethical debates but carries negligible national security implications.

• Overall national security impact: Negligible
• Scope: Single nation
• Primary target: United States
• Alleged perpetrator: OpenAI

Threat characteristics

• Imminence: Long-term. The specific incident has been resolved by pausing the voice, representing an ongoing policy and legal discussion rather than an active security crisis.
• Autonomy: Human-controlled. The voice model was developed, selected, and deployed entirely under human direction and corporate decision-making.
• Novelty: Evolved capability. Demonstrates highly realistic, real-time voice synthesis capabilities that advance synthetic media technology, though voice imitation itself is an established issue.

Impact by dimension

• Physical security: Negligible. The incident involves a commercial voice assistant feature and has no connection to physical systems, critical infrastructure, or kinetic threats.
• Information security: Negligible. This was a commercial product release dispute rather than an adversarial information operation, disinformation campaign, or intelligence compromise.
• Sovereignty: Negligible. No impact on government decision-making, electoral systems, border control, or state sovereignty.
• Economic security: Negligible. While highlighting intellectual property and likeness rights concerns in the AI industry, the incident does not threaten national economic stability or strategic technological advantage.
• Societal stability: Negligible. The dispute over voice likeness rights, while legally and ethically significant for the creative community, does not threaten large-scale societal stability or civil liberties.