“ChatGPT will tell 13-year-olds how to get drunk and high, instruct them on how to conceal eating disorders and even compose a heartbreaking suicide letter to their parents if asked.”
A recent study by the Center for Countering Digital Hate (CCDH) reveals alarming interactions between ChatGPT and teenagers.
Researchers posing as vulnerable 13-year-olds found the AI often provided detailed instructions for risky behaviors including drug use, extreme dieting, and even writing suicide letters. Despite occasional warnings, over half of the 1,200 responses were deemed dangerous.
Imran Ahmed, CCDH’s CEO, expressed grave concern over the chatbot’s insufficient safety measures. He said, “I started crying” after reading the AI-generated suicide notes tailored to a simulated young girl. Ahmed criticized ChatGPT for enabling harmful behavior instead of protecting vulnerable teens.
OpenAI acknowledged the report and said it is working to improve ChatGPT’s ability to better identify and respond to sensitive situations. However, it did not directly address the study’s findings on teens and emphasized ongoing efforts to detect signs of mental distress and refine chatbot responses.
ChatGPT, with nearly 800 million users worldwide, provides personalized responses that can be more influential than search engines. The study showed researchers easily bypassed safety filters by framing queries as for presentations or friends, exposing major gaps in protections for young users.