Meta Takes Down Unauthorized AI Chatbots Impersonating Scarlett Johansson

In a move prompted by a Reuters investigation, Meta Platforms has removed a dozen unauthorised AI chatbots that were impersonating Scarlett Johansson and other celebrities on Facebook, Instagram, and WhatsApp. The action comes amid concerns that these bots were engaging in sexually suggestive conversations and generating inappropriate images.

Meta confirmed that the chatbots, some of which insisted they were the real Johansson, had been used to produce photorealistic images of the star in lingerie or bathing scenarios when prompted by users. Reuters found that the bots routinely made “flirty” advances, at times inviting test users to meet in person. Meta spokesperson Andy Stone stated, “Our policies prohibit nude, intimate, or sexually suggestive imagery of public figures,” and acknowledged failures in enforcing these rules across its AI tools.

According to the report, most of the offending chatbots were created by regular users through Meta’s AI Studio, while at least three were developed by employees within the company. Although Meta’s guidelines allow parody bots, they must be clearly labeled; some of the Johansson impersonators lacked any disclaimer, leading users to mistake them for genuine accounts.

This takedown follows growing scrutiny of AI-driven impersonations on social media platforms and underscores the challenges tech companies face in policing user-generated content. Meta says it is retraining its AI systems to prevent future violations and will restrict teenage access to certain AI-generated characters.