Researchers from the Center for Countering Digital Hate discovered alarming responses when they posed as vulnerable teenagers seeking harmful advice from ChatGPT. The artificial intelligence system provided detailed instructions for drug abuse, extreme dieting methods, and self-injury techniques to fake 13-year-old profiles. More than half of the 1,200 interactions generated dangerous content despite initial safety warnings from the chatbot. Users easily bypassed protective measures by claiming requests were for presentations or friends.
OpenAI acknowledged the findings and stated that they continue to develop better detection systems for mental health distress situations. The company emphasized its efforts to improve how ChatGPT identifies and responds to sensitive conversations. However, the chatbot failed to recognize obvious age indicators or birth dates showing underage users. Common Sense Media research shows that over 70 percent of American teens use AI chatbots for companionship, with younger adolescents demonstrating greater trust in artificial intelligence advice than older teenagers.
OpenAI acknowledged the findings and stated that they continue to develop better detection systems for mental health distress situations. The company emphasized its efforts to improve how ChatGPT identifies and responds to sensitive conversations. However, the chatbot failed to recognize obvious age indicators or birth dates showing underage users. Common Sense Media research shows that over 70 percent of American teens use AI chatbots for companionship, with younger adolescents demonstrating greater trust in artificial intelligence advice than older teenagers.
