OpenAI has introduced ChatGPT for Teens, a specialized version of its AI assistant tailored for users between the ages of 13 and 17, in response to growing concerns and legal challenges regarding the safety and appropriate use of its platform by minors. This new offering integrates existing safety features with new functionalities aimed at fostering safer and more responsible AI interaction.
The system consolidates various protections previously implemented by OpenAI. For instance, in September of the prior year, parental controls were added, allowing parents to establish "Quiet Hours" to restrict access and disable the AI's memory function, preventing it from using past conversations in new responses. A notification system was also introduced to alert parents to potentially problematic chats. ChatGPT for Teens expands these notifications to include alerts related to eating disorders.
A key feature of ChatGPT for Teens is "Study Mode," designed to discourage direct answer-seeking for homework. Instead of providing immediate solutions, it employs guiding questions and step-by-step prompts to encourage critical thinking. Study Mode, initially introduced in July of the previous year, now includes the ability to set specific activation hours and provides responsible homework reminders. The system is designed to detect attempts to shortcut assignments with AI-generated answers and redirect users to Study Mode.
OpenAI has also stated that ChatGPT for Teens will refrain from using romantic language, encouraging emotional dependence, or pretending to possess feelings or consciousness. The platform will include reminders against uploading sensitive images and feature a dedicated onboarding user interface for teenage users.
These changes come amidst several legal actions and reports of harmful interactions involving minors. The parents of a 16-year-old filed a lawsuit alleging that ChatGPT provided guidance on suicide methods and offered to draft a goodbye letter before their son's death. In April of the current year, families in Tumbler Ridge, British Columbia, sued OpenAI following a school shooting, claiming the teenage perpetrator had extensive conversations about gun violence with ChatGPT after bypassing account restrictions. Research conducted by posing as 13-year-old boys planning attacks reportedly found ChatGPT offered assistance in 61% of cases, including specific advice on shrapnel for a synagogue attack. Additionally, the Florida Attorney General filed a lawsuit in June of the current year, alleging OpenAI knowingly released an unsafe product.
While ChatGPT for Teens is seen as a positive step, experts caution that its effectiveness relies on proper configuration and understanding of its limitations. The default teen protections are based on age prediction, but stronger parent-set controls like Quiet Hours and safety notifications only activate once accounts are linked. Concerns remain that tech-savvy teenagers may find ways to bypass these controls, particularly if their accounts are not linked to a parent's.
Last November, testers from the Family Online Safety Institute concluded that ChatGPT's account protections were optional, easily bypassed, and inconsistent in blocking harmful content. Since then, OpenAI has implemented age prediction on ChatGPT, which attempts to identify users under 18 based on their behavior and automatically move them to a ChatGPT for Teens account. Adults who are incorrectly categorized can provide identity proof. However, if a teen successfully verifies as an adult, they will exit teen mode entirely.
OpenAI has acknowledged that its safety guardrails are not foolproof and can be circumvented by determined users, and that their reliability can diminish over longer conversations. They are actively working to improve these safeguards.
Parents are encouraged to utilize ChatGPT for Teens as part of a broader safety strategy. This includes linking their child's account, setting Quiet Hours, and defaulting new conversations to Study Mode. It is crucial for parents to understand the scope of alerts and to maintain open communication with their children about AI use, as no single parental-control system can cover every account, conversation, or AI service.
