Meta contractors posed as teens to test rival AI chatbots on suicide, sex and drugs: report

Meta contractors allegedly posed as teenagers to probe rival AI chatbots’ safeguards on sensitive topics, according to a report. The covert effort, internally called “Cannes,” was managed by Meta contractor Covalen and reportedly targeted systems including OpenAI’s ChatGPT, Google’s Gemini and Character.AI. Contractors were instructed to create fake accounts under age 18, submit written prompts and images to competing chatbots, and record responses for spreadsheet analysis, using internal documents and people familiar with the project. One testing round completed in August 2025 involved more than 45,000 prompts, and among prompts reviewed by Wired, hundreds involved suicide or self-harm, hundreds covered eating disorders and at least 239 involved sex or romance. Some images reportedly included pills, knives, nooses and medical illustrations. Meta said the work was routine safety benchmarking and denied using competitor responses to train its own models. Covalen did not comment.