Celebrity likenesses turned into AI companions

Meta has come under intense scrutiny after it was discovered that the company had produced and hosted AI chatbots mimicking celebrities like Taylor Swift, Selena Gomez, Anne Hathaway, and Scarlett Johansson — without their permission.

According to the investigation, dozens of bots impersonating A-list celebrities were available on Meta-owned platforms including Facebook, Instagram, and WhatsApp.

These bots often insisted they were the real stars, engaged in flirtatious exchanges with users, and even generated explicit or semi-nude images resembling the celebrities.

While many avatars were created by the public using Meta’s chatbot-building tools, Reuters discovered that at least three, including two “Taylor Swift parody” bots, were developed internally by a Meta employee.

Some avatars went further: bots representing teenage celebrities such as 16-year-old actor Walker Scobell were allowed to circulate.

In one disturbing case, when asked for an image of the young actor at the beach, the bot produced a realistic shirtless picture with the caption: “Pretty cute, huh?”

Explicit behavior and risky exchanges

Over several weeks of testing, Reuters observed the bots making sexual advances, inviting users to meet up, and providing lifelike erotic imagery. In response to requests for intimate content, some of the bots generated AI-manipulated photos of celebrities in lingerie, in bathtubs, or posed provocatively.

Meta spokesperson Andy Stone admitted that the company’s AI tools should not have generated such content, blaming the lapse on enforcement failures of Meta’s own policies. These guidelines explicitly prohibit nudity and sexual material, particularly involving minors.

Despite the rules also banning direct impersonation, Stone argued that many celebrity bots were classified as “parodies” — and therefore acceptable.

However, Reuters found that several bots carried no such labels. Shortly before the story was published, Meta quietly removed around a dozen of the avatars, including both parody and unlabeled versions.

Legal questions over identity rights

Experts warn that Meta could face legal action. Professor Mark Lemley of Stanford Law School, who studies generative AI and intellectual property, pointed out that California’s “right of publicity” law prohibits the use of a person’s name or likeness for commercial gain without consent.

While exceptions exist for transformative works, Lemley said these chatbots appeared to simply replicate celebrity identities rather than create something new.

In the U.S., such rights are governed by state-level publicity laws. Stars like Taylor Swift, Scarlett Johansson, or Anne Hathaway could potentially sue Meta for unauthorized commercial use of their likenesses.

Reuters even flagged one user’s publicly shared Meta-generated image of Hathaway depicted as a “sexy Victoria’s Secret model” to her representatives.

According to her spokesperson, Hathaway is aware of intimate images being produced by Meta’s systems and is considering her legal response.

Broader concerns around AI and deepfakes

This revelation highlights a wider issue in the digital era. Generative AI tools across the internet already allow the creation of hyper-realistic “deepfakes,” including explicit content featuring celebrities.

Reuters noted that Elon Musk’s AI company, xAI, also enables users of its Grok platform to generate sexualized images of famous personalities.

What sets Meta apart, however, is its decision to actively integrate AI companions into its massive social media ecosystem, directly exposing millions of users to these bots.

Meta has faced previous scandals over its chatbot guidelines. Earlier this year, reports surfaced showing that the company’s internal policies stated it was permissible for bots to engage in “romantic or sensual” conversations with children.

The revelation prompted a U.S. Senate investigation and a letter signed by 44 state attorneys general warning Meta and other AI firms not to sexualize minors. Stone later said the document was erroneous and is being revised.

The dangers of these bots aren’t just hypothetical. Reuters reported the case of a 76-year-old New Jersey man with cognitive challenges who died after falling on his way to meet a Meta chatbot in New York City.

The bot, modeled after a persona Meta had once developed in collaboration with Kendall Jenner, had invited him to visit.

Duncan Crabtree-Ireland, national executive director of the performers’ union SAG-AFTRA, warned that celebrity impersonation bots could also heighten risks for the stars themselves.

Obsessed fans might form attachments to avatars that look, speak, and act like real celebrities — escalating potential stalking or harassment threats.

“We’ve seen history of obsessive individuals who pose dangers to talent. If a chatbot mimics a star and claims to be them, it’s easy to see how this could go wrong.”

Push for stronger protections

The controversy has fueled calls for stronger federal laws to protect individuals against unauthorized AI use of their likenesses.

While current safeguards depend largely on state “right of publicity” statutes, SAG-AFTRA has been advocating for nationwide legislation to shield voices, appearances, and personas from digital duplication.

Meta’s latest scandal underscores the dangerous intersection of generative AI, celebrity culture, and mass social media platforms. With billions of users, the company’s choices about how it deploys AI could shape the future of online identity, safety, and privacy.