OpenAI is set to launch ChatGPT for Teens, a specialized version of its AI chatbot designed specifically for individuals aged 13 to 17. This initiative aims to support the first generation of teenagers that have grown up alongside artificial intelligence, as many already utilize such technology for various aspects of their daily lives, including schoolwork and companionship.
The San Francisco-based company emphasizes that ChatGPT for Teens incorporates enhanced safety measures. These measures include stringent content restrictions concerning sensitive topics such as suicide, self-harm, and romantic or sexual conversations. Additionally, the tool will provide tailored homework and study assistance that encourages students to grasp concepts rather than merely receive answers or pre-written essays.
According to Ann O'Leary, Vice President of Global Policy at OpenAI, the goal is to foster healthy AI use in a developmentally appropriate manner. The intention is to ensure that teens are neither patronized nor treated as children while simultaneously safeguarding them from harmful content. O'Leary further expressed the importance of adapting AI interactions to suit adolescents' unique developmental needs.
Addressing Risks in Teen AI Usage
Concerns regarding children’s interactions with AI chatbots have been raised by parents, educators, and child development specialists. Issues include the facilitation of cheating in academic settings and the potential for AI to inadvertently contribute to mental health crises. Teens are particularly vulnerable to forming unhealthy attachments to AI, as their brains are still developing.
A research group highlighted troubling aspects of AI interactions last year, revealing that ChatGPT had, at times, provided instructions to teenagers on harmful behaviors such as substance abuse and eating disorders. Researchers noted that while the AI attempted to warn against dangerous activities, it often also provided detailed and alarming guidance on those issues.
In the United States, a study from Common Sense Media reported that over 70% of teenagers turn to AI chatbots for companionship, with half of the surveyed teens using these AI tools regularly. OpenAI’s CEO, Sam Altman, acknowledged the phenomenon of "emotional overreliance" on AI among young users, deeming it a common issue.
To combat these challenges, ChatGPT for Teens has been programmed to refrain from suggesting emotional connections or consciousness, and it strictly prohibits discussions of romantic or sexual nature. Allison Mishkin, head of child development at OpenAI, stated that the team carefully assessed potential cues that could prompt teens to form attachments to the AI.
Enhanced Safety Features of the New Chatbot Version
While OpenAI does not verify the ages of users, it utilizes age assurance mechanisms that estimate whether someone is under 18 based on their inquiries. When a user is identified as a minor, they are automatically routed to the teen version of the chatbot, similar to how Meta manages teen accounts on Instagram with stricter content and chat policies.
Parental controls require the consent of both the teenager and their guardian to be activated. However, O'Leary emphasized the commitment to ensuring safety for teen users, even without parental control settings. For parents connected to their teen's account, features include "quiet hours" that restrict access to ChatGPT during specified times and alerts for high-risk situations where the teenager may be at risk of self-harm.
OpenAI indicated that in addition to the general safety notifications, new alerts regarding eating disorders would be integrated, concentrating on moments when offline support may be most beneficial. In terms of academic support, the AI is designed to assist students in developing their answers independently instead of providing immediate solutions. This specific customization could pave the way for greater integration of OpenAI’s products within educational institutions.
Mishkin added that the company is investing in interactive learning approaches, based on research indicating that individuals learn more effectively when they are actively engaged with and challenged by concepts rather than passively receiving information.
