The history of celebrity likeness disputes is long and often absurd — Bette Midler once sued Ford Motor Company over a sound-alike singer in a commercial, and won — but the latest chapter may be the strangest yet.
Tana Mongeau, a creator with followings in the millions across Instagram, YouTube and TikTok, took to X this month to express alarm at a video from Miso Labs, an AI voice generation company. "Why is this like my voice. Help," Ms. Mongeau wrote, reacting to what she felt was a digital doppelgänger built without her consent.
The company's co-founder, Aoden Teo, had previously boasted that Miso could "clone any voice with just 10 seconds of audio," and acknowledged that another model was based on the creator Salman Khan, adding, with what one assumes was meant to be disarming candor, "Pls don't sue us."
(One suspects the legal department may have preferred a different phrasing.)
The episode illustrates a broader anxiety now coursing through the creator economy: generative AI tools have become sophisticated enough to mimic a person's look and voice from mere seconds of publicly available content. For creators, whose careers are built on the distinctiveness of their presence, and for brands, who pay handsomely for that distinctiveness, the implications are considerable.
Some creators have attempted to get ahead of the problem by building authorized versions of themselves. Khaby Lame, the TikTok star with nearly 162 million followers, signed a $975 million deal earlier this year to license his AI-generated digital twin — though the share price of his partner company has since fallen by more than 90 percent, and Mr. Lame has said little about the arrangement. Another creator, Vicky Waldrip, released an AI clone last year that has since been "discontinued," according to a spokesperson.
Legal protections remain thin. Frank Poe, an attorney who works with creators, noted that trademarking specific phrases or images — as Taylor Swift and Matthew McConaughey have done — offers some defense, but proving that a company knowingly used someone's content to train a model requires costly and time-consuming litigation.
"It's very difficult to prove," Mr. Poe said.
For now, the creator economy finds itself in a familiar position: waiting for the law to catch up with the technology, and hoping the technology doesn't do too much damage in the meantime.
Original story published in Digiday: "The Rundown: AI clones split the creator economy"