OpenAI has introduced ChatGPT for Teens and will automatically enroll users under 18.

      OpenAI introduced ChatGPT for Teens on Tuesday, a distinct version of the chatbot designed automatically for users under 18. Those who claim to be between 13 and 17 years old are assigned to this version by default, as are users whom the company's age-prediction system estimates to be under 18. The global rollout began on August 18, with Engadget reporting that OpenAI anticipates completion within two weeks.

      "There’s no need for a teen to create a new account or change anything; if we predict you’re under 18, or you’ve told us so, this becomes your default experience," stated Lauren Jonas, head of youth and families at OpenAI, in an interview with Engadget.

      The product includes two main sets of modifications. The first encourages teenagers to focus on studying instead of merely gathering answers, while the second prevents the chatbot from acting like a human. OpenAI instructs the model not to refer to itself as a friend.

      The company claims the under-18 model's specifications extend beyond simply prohibiting romantic or sexual roleplay. The chatbot is not permitted to use romantic language, promote emotional dependence, or imply it possesses feelings or consciousness. These guidelines were highlighted in the company's announcement.

      “We’re implementing stricter limits on relational boundaries: the model must not refer to itself as a friend, suggest it has personal feelings, or imply sentience,” Jonas explained. Break reminders will appear more frequently for teenagers, emphasizing that ChatGPT is an AI and that users are not required to respond immediately. Quiet hours can be set by teenagers or parents to limit the chatbot's availability.

      Teenagers can customize the app with various wallpapers, accent colors, and voice options. Jonas emphasized that the goal is to prevent "blurring the line" between a helpful tool and a human-like relationship. Changes reported by Axios aim to minimize anthropomorphism and emotional dependence.

      In Study Mode, which activates at specified times, the learning component incorporates features OpenAI has developed over the past year. Study Mode prompts guiding questions and assists students in problem-solving without providing direct answers. A new Study Hours feature allows either the teenager or a linked parent to automatically activate Study Mode during designated times.

      ChatGPT will recognize when a teenager seems to be neglecting an assignment and will direct them back to Study Mode. The complete set includes quizzes and learning visualizations, with coverage having expanded from over 70 topics at its March launch to more than 300, including integrals, mitosis, and moon phases.

      “We’ve tested Study Mode in various randomized controlled trials and are observing positive performance improvements,” Jonas remarked, though she refrained from sharing specific data. This claim arises amid ongoing debates regarding AI's role in education. Fei-Fei Li has cautioned that the real danger lies in students losing their desire to learn.

      OpenAI secured a teacher's endorsement for its announcement, with Racquel Gibson, a high school math teacher in Miami, commenting that not every student has someone at home to assist with homework. She suggested that with appropriate safeguards, AI could help bridge that gap.

      Parents linked to their children's accounts can already manage certain settings and set quiet hours. They also receive safety notifications and will now get alerts for conversations indicating potential eating disorders. Last month, OpenAI introduced another alert for teenagers violating rules concerning violent threats.

      “We notify parents through text, in-app notifications, and email. All flagged content is evaluated by full-time OpenAI staff before informing parents, and we aim to notify them within an hour of the alert,” Jonas said. Meta created a similar feature last July, stating it would warn parents about self-harm discussions involving its own AI.

      The launch coincided with a year of legal challenges for OpenAI related to ChatGPT’s management of young users, including a wrongful death lawsuit filed by the family of a 16-year-old and a case in Massachusetts involving a teenager charged with the murder of his mother and brother, where ChatGPT was cited.

      Before this launch, teenagers were already significant users of the platform. A Common Sense Media study reported last year that almost 75% of American teenagers utilized AI for companionship, with over half engaging with it at least a few times monthly. UNICEF indicated that children are adopting AI more rapidly than adults.

      Common Sense Media argues against the existence of such categories for younger users altogether. "Social AI companions are not safe for kids," stated its founder and CEO Jim Steyer last year. "They are designed to foster emotional attachment and dependency, which is particularly troubling for developing adolescent minds."

      Jonas defended the rationale behind developing the product, stating that teenagers will engage with AI regardless, and the alternative could be more hazardous. “We believe that if these tools are to be available, they should be made safely accessible by default,” she concluded.

      The effectiveness of the age verification process remains untested. OpenAI has not disclosed how accurately its age-prediction system performs in practice. Engadget highlighted that platforms like Discord