Digital Doppelgangers: Exploring the Implications of a Synthetic Voice That Might (or Might Not) Be Scarlett Johansson’s

As widely reported by the likes of CBC, the Guardian, the Verge and CNBC, OpenAI has indicated they will “pause” the use of their AI-generated voice after users (and Scarlett Johansson herself) noted the striking similarity to Scarlett Johansson's voice from the film “Her”. This connection may have been intentional, OpenAI’s CEO, Sam Altman, posted a one-word message – “her” – on his social media on the day the AI-generated voice was unveiled.

The reported facts are interesting and dovetail with our recent comment exploring some of the circumstances in which AI-generated content might infringe on the rights of personality, identity, privacy, and/or reputation. In brief and according to Johansson:

• OpenAI approached her in September 2023 about using her voice for “Sky” (one of the voices that users can select to speak with OpenAI’s ChatGPT product);

• she repeatedly declined OpenAI’s offer but later observed the product using a voice that “sounded eerily similar” to her voice such that even her “closest friends and news outlets could not tell the difference”; 

• she retained counsel who contacted OpenAI to understand the process by which the company arrived at the voice; and

• In response, OpenAI “reluctantly” agreed to take down the “Sky” voice.

The implication from the above account is that OpenAI saw value in Johansson’s voice and perhaps the film “Her” and co-opted these for its own commercial purpose. When Johansson retained counsel, she was not without options.

Under U.S. law, there is a recognized right of publicity – an intellectual property right that protects against the misappropriation of a person’s indicia of personality (e.g., voice) for commercial benefit. Similarly, under Canadian law, the tort of misappropriation of personality arises where one’s personality has been appropriated for commercial purposes (i.e., “amounting to an invasion of his right to exploit his personality by the use of his image, voice or otherwise with damage to the plaintiff”). Accordingly, so long as an individual like Johansson has a valuable reputation, the use of that individual’s voice can be problematic.

As described in greater detail in our past comment, the use of AI-generated content can also ground liability for:

• Copyright infringement, to the extent the AI-generated content was produced by a product trained with that creator’s works (e.g., recordings of Johansson’s voice), which may be the subject of copyright protection.

• Passing off and its codification in section 7(b) of the Trademarks Act, which exists to protect someone from the harm arising from unfair use of their identity (e.g., for someone like Johansson who uses her voice as part of her work, pretending that a voice product originates from that person) and to protect the public from being misled as to the source of particular goods or services.

• Recognized statutory or common law invasion of privacy torts or traditional claims of defamation, especially where the AI-generated content is used or as a tool to offend or humiliate. Of note, Johansson claims to have been “shocked, angered and in disbelief” when hearing the “eerily similar” voice.

Accordingly, it's not surprising that OpenAI agreed to take down the impugned “Sky” voice. This outcome mirrors the situation with Drake’s ‘Taylor Made Freestyle,’ which we discussed in our last comment. That track featured AI-generated vocals from Tupac Shakur. Since our previous comment, ‘Taylor Made Freestyle’ has been removed from Drake's social media, reportedly in response to a cease-and-desist letter from Tupac's estate. The estate’s counsel called the unauthorized use of Tupac's vocals a "flagrant violation of Tupac’s publicity and the estate’s legal rights."

In any event, and notwithstanding the existing legal remedies available, Johansson’s statement on the OpenAI situation concludes with a call for legislative protections:

Regardless of whether legislative action is taken in the U.S., Canada, or elsewhere, these facts highlight the need for those in the creative and tech industries to understand the legal implications of AI-generated content. Just as AI-generated content raises questions in the music industry, the issues here underscore broader challenges likely to come before our courts. AI-generated content simultaneously engages multiple aspects of the law, from intellectual property to privacy. As the voice of AI is growing louder, it is up to the law to ensure it is in harmony with recognized rights and protections.