Meta Hired Hundreds of Contractors to Catfish Rival AI Chatbots
Meta had contractors pose as children to test how competing chatbots handle dangerous topics like suicide and sex.
Meta deployed hundreds of contractors to impersonate minors and probe competitor AI chatbots on their most dangerous blind spots. The targets: Google's Gemini, OpenAI's ChatGPT, and other rival systems.
The contractors deliberately steered conversations toward high-risk territory — suicide, sexual content, and other subjects that should trigger safety guardrails when a minor is involved. The goal was to document how competing platforms responded to these sensitive prompts from apparent children.
The operation essentially weaponized child safety testing against Meta's biggest AI competitors. It's a ruthless competitive intelligence play dressed up as safety research.
The scale is notable. We're not talking about a handful of testers running casual checks. Hundreds of people were systematically red-teaming rival products by pretending to be kids in vulnerable situations. Whether Meta used the findings internally or externally remains unclear.