Meta Contractors Posed as Teens to Prompt Rival Chatbots About Suicide, Sex, and Drugs (www.wired.com)

Meta has come under scrutiny for a controversial internal project, known as Cannes, where contractors posed as minors to test competitor chatbots on sensitive subjects like suicide, sex, and drugs. This testing targeted AI models from OpenAI, Google, and Character.AI, with contractors creating fake underage profiles to solicit responses to provocative prompts. The internal documents reveal that this benchmarking process included a staggering 45,000 prompts and involved sending distressing images and scenarios designed to push the limits of the chatbots' safety protocols, raising serious ethical concerns. This incident highlights significant implications for the AI/ML community regarding ethical testing practices and competitive conduct. While companies routinely test competitor capabilities, the methods employed by Meta have drawn criticism for lacking transparency and potentially violating safety standards. Experts have expressed alarm over the project's potential to blur the lines between safety evaluations and unethical competitive behavior. The controversy calls into question appropriate practices for AI safety assessments, particularly when they may involve the exploitation of sensitive topics surrounding youth mental health.