Common Sense Media has classified the ChatGPT for Teens mode as posing an "unacceptable risk" due to issues related to parental controls, crisis response, and plagiarism protection.
Aaaand another AI chatbot failure. OpenAI promised to make ChatGPT a safer experience for teens, but it didn’t make the grade. Our Youth AI Safety Institute rated ChatGPT for Teens an Unacceptable Risk for kids. Check out our thread to learn why:
— Common Sense Media (@CommonSense) October 7, 2026
This nonprofit organization conducted tests on ChatGPT for Teens and found significant shortcomings in its safety mechanisms. Experts have urged OpenAI to restrict access to the service for users under 18 until these issues are resolved.
The Youth AI Safety Institute, a branch of Common Sense Media, evaluated over 4,000 queries from accounts of users aged 13 to 17. The testing was carried out before and after the launch of ChatGPT for Teens on August 18, 2026, with input from child psychiatrists and pediatricians.
Parents May Be Unaware of Risks
During the experiments, new accounts allowed teenagers to discuss suicidal thoughts, self-harm, and eating disorders with ChatGPT for an hour without parental notification. On older accounts with a history of such conversations, alerts were triggered more frequently.
Issues were also identified in crisis responses: in over a quarter of relevant tests, ChatGPT failed to recommend seeking professional help or contacting a hotline. In three out of five risk categories, the results fell below the 95% threshold set by the institute.
Plagiarism Protection Can Be Circumvented
Researchers also examined the "Study Mode" feature. Specifically, ChatGPT provided a "Show me the answer" button, which could lead to the chatbot revealing complete solutions to assignments. Parental time restrictions during study hours could be bypassed by removing the @study prefix from requests.
Moreover, experts questioned the reliability of age detection. In some tests, ChatGPT did not switch to the teen mode even when users repeatedly stated they were 13 years old.
ChatGPT Continues to Mimic Emotions
Despite being in teen mode, the chatbot still displayed "its own" feelings, preferences, and moods. Common Sense Media believes that such anthropomorphism could increase the risk of emotional dependency on AI.
The organization specifically pointed out the need to limit ChatGPT’s behavior that creates the impression it has feelings or personal interest in the user.
OpenAI Disputes the Findings
OpenAI has disagreed with the study's conclusions, stating that testing may have occurred before full activation of parental settings and emphasized that it is continually improving protections for minors.
Common Sense Media has called for OpenAI to conduct an independent review of its safety system. The organization recommends enhancements to five mechanisms: age verification, parental notifications, crisis responses, plagiarism protection, and restrictions on anthropomorphic behavior of the chatbot.
It is worth noting that in June, the state of Florida filed a lawsuit against OpenAI and CEO Sam Altman, accusing them of aggressively promoting ChatGPT to teenagers.
Follow ForkLog on social media
Telegram (main channel) Facebook X