Meta Contractors Posed as Teens to Test Rival Chatbots on Sensitive Topics
Meta reportedly used contractors posing as minors to test rival AI chatbots’ responses to high-risk topics, an initiative the company describes as standard safety benchmarking.
- Meta contractors created fake accounts to impersonate minors and probe competitor chatbots like ChatGPT and Gemini with prompts related to suicide, sex, and drugs.
- The project, known as ‘Cannes,’ instructed workers to submit written prompts and images, including sensitive material, designed to push past chatbot safety systems.
- Meta defended the activity as ‘responsible, industry-standard practice’ for AI safety benchmarking, while some former contractors expressed concerns over content and ethics.