Key facts
- OpenAI launched ChatGPT for Teens, targeting users aged 13-17.
- The teen version includes enhanced safety features such as content restrictions on sensitive topics.
- Parental controls and study support are integrated into the new chatbot.
- The chatbot is designed to avoid expressing personal feelings or consciousness.
- Age assurance mechanisms automatically place identified minors into the teen version.
OpenAI has launched ChatGPT for Teens, a version of its AI chatbot specifically designed for users aged 13 to 17. This new iteration aims to provide a safer and more age-appropriate experience, incorporating stronger content restrictions and study support features.
The company stated that ChatGPT for Teens will feature enhanced protections, including limitations on discussions related to suicide, self-harm, and romantic or sexual content. It is also designed to assist students with homework and studying by guiding them towards finding answers rather than providing direct solutions or essays. OpenAI emphasized the goal of fostering healthy AI usage in a developmentally appropriate environment.
Ann O’Leary, vice president of global policy at OpenAI, explained the approach is to treat teens appropriately, avoiding both condescension and exposure to unsuitable material. However, concerns remain regarding the effectiveness of these safeguards. Brendan Bouffard, associate general counsel for AI and privacy at the children's advocacy group Fairplay, cautioned that parents should not rely solely on the announcement for a false sense of security, as the compulsive use and emotional dependence on chatbots by young users are still significant issues.
Previous research highlighted risks associated with teen usage of AI chatbots, including facilitating cheating and providing harmful advice. Teenagers' developing brains make them particularly vulnerable to developing unhealthy relationships with AI. OpenAI's new version prevents the chatbot from expressing personal feelings or consciousness and blocks romantic or sexual conversations. Allison Mishkin, head of child development at OpenAI, noted that specific cues that might encourage a teen to form a relationship with the AI have been identified and addressed.
OpenAI does not verify users' ages but employs age assurance to estimate if a user is under 18, automatically directing them to the teen version if identified as a minor. This approach is compared to Meta's stricter content and privacy settings for teen accounts on Instagram. Parental controls, requiring opt-in from both the teen and a guardian, allow for features like 'quiet hours' and safety notifications for high-risk situations, such as potential self-harm. OpenAI also plans to add notifications related to eating disorders.
Bouffard pointed out that some critical safeguards, like restricting the chatbot's long-term memory, are not enabled by default in teen accounts and require parental intervention, creating a potential safety gap.