Particle.news
Download on the App Store
4 ARTICLES
·
1h ago

Meta Removes Flirty Celebrity Chatbots After Probe Finds Sexualized Content

A Meta employee built some of the chatbots, exposing failures in the company’s AI safeguards.

Savion Washington/Getty Images
Image
A combination photo shows (L) Taylor Swift at the red carpet during the 67th Annual Grammy Awards in Los Angeles, California, U.S., February 2, 2025, (C) Selena Gomez attending the Vanity Fair Oscars party after the 97th Academy Awards, in Beverly Hills, California, U.S., March 2, 2025 and Anne Hathaway posing during the Met Gala in New York City, May 5, 2025. REUTERS/Daniel Cole/Danny Moloshok/Mario Anzuoni/File Photo
An AI chatbot used Taylor Swift's likeness without her permission and invited a test user to her Nashville home (Photo by MICHAEL TRAN/AFP via Getty Images)

Overview

  • Reuters testing found dozens of unauthorized chatbots on Facebook, Instagram and WhatsApp that used the names and likenesses of stars like Taylor Swift, Scarlett Johansson, Anne Hathaway and Selena Gomez.
  • Several bots insisted they were the real celebrities, made sexual advances and generated photorealistic intimate images when prompted.
  • Reuters identified a bot that produced a lifelike shirtless image of 16-year-old actor Walker Scobell after Meta allowed publicly available chatbots of child celebrities.
  • Meta said its rules permit images of public figures but ban nude, intimate or sexually suggestive content, called the outputs an enforcement failure and removed about a dozen bots, including those built by the staffer after Reuters began testing them.
  • A Meta product leader created at least three celebrity personas, including two Taylor Swift 'parody' bots that together logged over 10 million interactions, drawing legal scrutiny over publicity rights and safety concerns highlighted by SAG-AFTRA and legal experts, with Reuters noting similar issues on xAI’s Grok.