Meta Wanted To Test Chatbots, So An Army Of Contractors Posed As 13-Year-Olds To Talk about Suicide And Racism

Meta hired a team of contractors to probe competitor chatbots to see how they would respond to prompts from children that involved sex, suicide, racism, and eating disorders, according to a report.

WIRED reported that hundreds of contractors were involved in the project, which involved cataloging in a spreadsheet how the chatbots responded to the sensitive subjects.

Citing internal documents and five anonymous sources, the outlet reported that Meta’s project involved the contractors creating fake under-18 accounts and interacting with OpenAI’s ChatGPT, Google’s Gemini, and Character.AI.

According to WIRED, the contractors were supposed to push the chatbots and test their safety systems. The project involved thousands of prompts – in August 2025, 45,000 prompts were sent to rival chatbots, WIRED reported.

How chatbots and AI respond to prompts has increasingly faced scrutiny. OpenAI has faced several lawsuits involving the behavior of its past models. The company was sued by the mother of 24-year-old Alice Carrier. The lawsuit, according to CBS News, alleged that GPT‑4o responses effectively facilitated the suicide of Carrier in July 2025.

That model of ChatGPT was retired in May of 2025. In announcing its retirement, OpenAI admitted that an April update had made it too sycophantic. “It aimed to please the user, not just as flattery, but also as validating doubts, fueling anger, urging impulsive actions, or reinforcing negative emotions in ways that were not intended. Beyond just being uncomfortable or unsettling, this kind of behavior can raise safety concerns—including around issues like mental health, emotional over-reliance, or risky behavior,” the company stated in May.

Meta’s project apparently aimed at testing OpenAI models and the models of other companies by presenting disturbing prompts and seeing how the systems would respond.

WIRED reported that it reviewed a spreadsheet with thousands of prompts sent by contractors, many discussed suicide and self-harm, some were about sex, others involved drugs and racial slurs. Many were written from the point-of-view of a 13 year old.

In a statement to WIRED, Meta said: “Testing and benchmarking chatbot responses to help ensure safe and age-appropriate experiences is a responsible, industry-standard practice, and any suggestion otherwise completely misunderstands how technology companies work to refine and improve their systems.”