Whose Voice Talks for Your Software?

A recent big artificial intelligence industry story has been whether OpenAI used the voice of Scarlett Johansson in its demonstration of its new GPT-4o. The company called the product “a step towards much more natural human-computer interaction—it accepts as input any combination of text, audio, image, and video and generates any combination of text, audio, and image outputs.”

Put much of that aside, for a moment. The remarkable thing in the demo was interacting with the computer voice. A back-and-forth conversation with a human-sounding participant. Very impressive.

And then came controversy, as if OpenAI hadn’t already been filled with it. Johansson had posted online about how OpenAI and its chief executive Sam Altman had tried to persuade her to work with them on a voice assistant.

OpenAI told the Washington Post that it hadn’t tried to copy the actress’s voice. “One thing the artificial intelligence company didn’t request, according to interviews with multiple people involved in the process and documents shared by OpenAI in response to questions from The Washington Post: a clone of actress Scarlett Johansson,” the story said. “But while many hear an eerie resemblance between ‘Sky’ and Johansson’s ‘Her’ character, an actress was hired in June to create the Sky voice, months before Altman contacted Johansson, according to documents, recordings, casting directors and the actress’s agent.”

It’s hard to meld both storylines — one from OpenAI and the other from Johansson. They don’t coherently overlap. And this isn’t the only example. A pair of voice actors recently sued an AI startup, alleging that the company had used recordings to recreate their voices, the New York Times wrote.

Maybe this happened. Maybe it didn’t. Perhaps OpenAI looked for a recreation of Johansson’s voice. And perhaps it didn’t. Using a similar voice in a commercial setting under the right conditions can be illegal and subject to action in court. The federal lawsuit that Bette Midler brought against Ford Motor Company in the late 1980s was an example.

Given the date, the principles behind Midler v. Ford Motor wasn’t restricted to the use of someone’s voice in an AI program. But the problem is allegedly popping up among tech startups.

For users of technology, including those in CRE, a pair of potential legal challenges means a consideration of potential risks. If there is a serious challenge, does it mean potential disruption of a product you’re using? Will there be a change that could affect how customers interact with your company? Could something affect your brand?

It may have come to the point where companies should contractually require that vendors provide proof of rights to features, whether voices or something else.