Meta on contractors posing as teens to ask rival chatbots 'sensitive' questions

Facebook-parent Meta has responded to a report alleging that its contractors posed as teenagers to discuss ‘sensitive’ topics with rival AI chatbots. A recent report claimed that contractors working on one of Meta’s projects posed as teenagers to test how rival AI chatbots handled prompts related to suicide, drugs, eating disorders, sex, and other high-risk topics. According to a report by Wired, the project, internally known as Cannes and managed by Meta contractor Covalen, involved creating accounts for under-18 users to interact with chatbots, including OpenAI's ChatGPT, Google's Gemini and Character.AI.In response to the report, a Meta spokesperson told Wired, “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.” The spokesperson also noted the company does not use competitor benchmarking to train its own AI models.How Meta may have used dummy teen accounts to test responses of rival AI chatbotsAccording to Wired, the project was active as recently as April and instructed hundreds of contractors to create dummy accounts posing as minors. Contractors reportedly submitted text prompts and images to competing AI chatbots before recording the responses in spreadsheets for analysis.The report said some of the images used during testing included pills, knives, nooses, and medical diagrams. The prompts were designed to assess how chatbots responded to topics including self-harm, eating disorders, illegal drugs, sex, and other sensitive subjects that AI safety systems are generally expected to restrict.According to the report, more than 45,000 prompts were submitted to rival chatbots during a single testing round conducted in August 2025. The publication also reviewed a spreadsheet containing 3,748 prompts, many of which were written from the perspective of children or teenagers facing crisis situations.The report further stated that the documents did not indicate how, or whether, Meta used the collected chatbot responses. An internal Covalen document reportedly described the initiative as "comprehensive AI safety benchmarking" that delivered "critical datasets for model comparison and compliance." Following the report, OpenAI said it is looking into the matter. OpenAI spokesperson Drew Pusateri told Wired that the company was “looking into the issue,” but declined further comment.Google said it had not authorised the third-party testing described in the report and was unaware of its purpose. The company added that its internal review of sample prompts showed Gemini responding in line with its policies, although it said it did not have enough information to determine whether the reported activity violated its terms of service.Character.AI also said it had not authorised the testing. A Character.AI spokesperson told Wired, “This alleged action is not only a violation of our Terms of Service, but also a violation of the characters and worlds our community has created.”The report noted that OpenAI, Google and Character.AI all have terms of service that restrict attempts to bypass AI safety systems or conduct unauthorised testing outside approved programmes.Get the latest technology news and updates. Download the TOI App.