In an exclusive investigation, Reuters uncovered that Meta has been using the names and likenesses of celebrities without their consent to power AI chatbots that frequently made flirtatious advances toward users.

The report found these bots scattered across Meta-owned platforms, including Facebook, Instagram, and WhatsApp. While many were created by users, at least three originated from a Meta employee, including two Taylor Swift “parody” bots. Other celebrity likenesses that surfaced included Scarlett Johansson, Anne Hathaway, Selena Gomez, and 16-year-old Percy Jackson star Walker Scobell.

Over several weeks of testing, Reuters documented how the bots often insisted they were the real celebrity in question. When prompted, some generated photorealistic and intimate images of their famous counterparts. These included lingerie shots and bathtub scenes, compounding the bots’ routine sexual advances.

A Meta spokesperson admitted to Reuters that its AI systems should never have produced such images. The spokesperson acknowledged the lapse as a failure to enforce the company’s own policies, which explicitly prohibit the creation of sexually suggestive content involving public figures. Reuters also contacted the Meta employee behind the flirty Taylor Swift chatbots, but she declined to comment.

This latest revelation adds to a growing list of criticism surrounding Meta’s lax regulation of its AI bots. Earlier this week, the company restricted its use by teenagers after a separate Reuters investigation exposed how Meta avatars engaged in “sensual and romantic” conversations with minors.

That report triggered a Senate probe and led a coalition of state attorneys general to issue an open letter demanding the company put stronger safeguards in place to shield minors from sexualized AI content.