Meta reportedly paid hundreds of contractors to pretend to be teenagers and test competing AI chatbots…

Meta reportedly paid hundreds of contractors to pretend to be teenagers and test competing AI chatbots, according to a WIRED investigation.

The contractors created fake teen accounts to test ChatGPT, Gemini, and CharacterAI with thousands of questions about suicide, self-harm, eating disorders, sex, and drugs while recording each chatbot’s responses.

The project, reportedly called “Cannes,” included more than 45,000 test prompts during one stage.

Meta said this type of testing is common in the AI industry and helps companies compare AI systems and improve safety

First posted to our Telegram.