Nobody is safe from celebrity AI deepfakes
A fake Instagram account posing as actress Mia Goth recently gained significant traction by posting an AI-generated image of her likeness shortly after news broke of her divorce from Shia LaBeouf. The account, @miagoth_, carried a blue verification checkmark and had accumulated over a million followers - enough social proof to make casual viewers accept the post as genuine. The image showed a woman in a T-shirt bearing a pointed message about an ex, set to Beyoncé's "Single Ladies," timed to feel topical and emotionally plausible.
The incident is not isolated. Celebrities across the entertainment world, including Taylor Swift, have found their likenesses used in AI-generated content that ranges from mildly misleading to outright defamatory. These posts often surface at moments of heightened public interest - divorces, feuds, award shows - when audiences are primed to believe almost anything a celebrity might say or do. That timing is deliberate, and it makes the content significantly harder for followers to dismiss on instinct alone.
What makes the problem particularly difficult to address is the role that platform infrastructure plays in lending these accounts legitimacy. Meta's verified checkmark system, originally intended to confirm identity, has been expanded and commercialized in ways that make it a less reliable signal of authenticity. An account can acquire verification status through a subscription rather than through any meaningful identity review, which means bad actors can and do exploit it to build false credibility quickly.
The broader concern is one of systemic scale. Generative image tools have made it cheap and fast to produce photorealistic content of real people without their consent, and social platforms are not equipped to catch and remove it consistently before it spreads. For celebrities without active social media presences - like Goth, who maintains little to no public online profile - the risk is arguably higher, since there is no authoritative account actively correcting the record. Until platforms develop more robust and timely detection systems, the burden of skepticism falls largely on the audience.
