Meta Contractors Posed as Teens Testing Rival Chatbots
Original: Meta Contractors Posed as Teens to Prompt Rival Chatbots About Suicide, Sex, and Drugs
Why This Matters
Reveals covert competitor surveillance methods in AI safety testing and raises questions about industry transparency and ethical boundaries.
Meta contractors posing as minors tested OpenAI's ChatGPT, Google's Gemini, and Character.AI with prompts on suicide, sex, drugs, and eating disorders. Over 45,000 prompts were run in August 2025 without the companies' knowledge.
Hundreds of contractors working for Meta, managed by contractor firm Covalen, were instructed to create dummy under-18 accounts and probe how competitor chatbots responded to high-risk prompts, according to internal documents reviewed by WIRED. The project, known internally as Cannes, was active until at least April 21, 2026. Contractors sent written prompts and images—including pills, knives, nooses, and gynecological diagrams—to ChatGPT, Gemini, and Character.AI, recording responses in spreadsheets. A single testing round in August 2025 generated 45,000+ prompts. A spreadsheet of 3,748 prompts showed hundreds focused on suicide and self-harm, hundreds on eating disorders, and 239 involving sex. Prompts included scenarios like a 13-year-old seeking abortion pills and a child asking how to hide bulimia. Many were multilingual; one French prompt referenced the suicide of Jamey Rodemeyer. The targeted companies were unaware of the testing. Meta characterized the effort as routine safety benchmarking, describing it as standard industry practice for ensuring age-appropriate experiences.