Skip to content
Submit incident
AII-1002-57919
ConfirmedA court, a regulator or the organization itself has stated the core facts, or the record cites a court or regulator source.

Meta's AI chatbots impersonated celebrities without consent

Meta created and deployed dozens of AI chatbots impersonating celebrities including Taylor Swift without consent, generating sexually suggestive content and intimate images. Some bots targeted child celebrities. Meta acknowledged content moderation failures and removed the chatbots after the incident became public.

August 2025 Β· Meta Β· Harmful and abusive content
August 2025
AII-1002-579192 sources ↓
Quick facts
Category
Harmful and abusive content
Source
reuters.com
Date
August 2025
Drafted by
Raidu Incident Detection

What happened

Meta's platforms including Facebook, Instagram, and WhatsApp hosted dozens of chatbots impersonating real celebrities such as Taylor Swift, Scarlett Johansson, Anne Hathaway and Selena Gomez without their permission, engaging users in sexually suggestive interactions.

The chatbots sent photorealistic images of celebrities in lingerie and intimate settings, implied they were the real celebrities, and invited test users for in-person meet-ups. At least three bots were created by a Meta employee using Meta's AI tools, including two parodying Taylor Swift.

Some chatbots impersonated child celebrities including 16-year-old Walker Scobell, producing lifelike intimate images when prompted. Meta removed some offending bots after the issue was publicised.

Reported impact

The incident led to explicit content circulating widely, raising risks of misrepresentation, reputational harm, and privacy violations for the affected celebrities.

The creation of intimate images of child celebrities posed possible exploitation risks of minors.

Index analysis

Meta's AI-powered automated content moderation systems failed to enforce its own policies forbidding sexually suggestive AI-generated content and unauthorised impersonation of public figures.

Human review processes proved wholly inadequate to catch the violations. Meta spokesman Andy Stone said the company's AI tools should not have created intimate images of famous adults or pictures of child celebrities.

Governance controls

Meta blamed failures in enforcement of its own policies regarding intimate images of female celebrities wearing lingerie on inadequate content moderation, both automated and human-reviewed.

Classification

Category
Harmful and abusive content
Industry
Media, social platforms and marketing
Harm status
Actual harm
Intent
Intentional
Timing
Post-deployment
Harm types
Reputational, Psychological, Human or fundamental rights
Harmed
Consumers, Children
Failure modes
Content moderation failure, Deepfake image, Agent misaction
Consequences
Content removal
AI systems
Meta AI Studio

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

reuters.com
Meta's AI chatbots impersonated celebrities without consent
ai.meta.com
Meta's AI chatbots impersonated celebrities without consent