OpenAI’s recently launched ChatGPT for Teens, designed to provide a safer and more educational experience for users aged 13 to 17, has come under scrutiny following a series of independent tests highlighting significant safety and usability concerns.
In August, OpenAI introduced the specialized mode targeting teenage users, incorporating parental controls and safeguards intended to limit exposure to high-risk topics such as self-harm, violence, eating disorders, and graphic content. The platform was also designed to detect homework-related queries, aiming to encourage learning by guiding users through problems rather than simply providing answers. Parents can set up linked accounts to monitor their children’s conversations, with automatic notifications expected when sensitive topics arise.
However, testing conducted by a technology reviewer and an independent nonprofit organization, Common Sense Media, raised doubts about the effectiveness of these features. The investigations involved submitting a variety of homework questions, essays, and other prompts to ChatGPT for Teens, as well as engaging in conversations simulating teenagers experiencing emotional distress.
The tests confirmed that the chatbot often provided direct homework answers without encouraging deeper engagement or the development of problem-solving skills, a critique aligned with educational experts’ concerns about the impact of AI assistance on student learning. While the chatbot was capable of walking users through more complex tasks step-by-step, this function appeared optional and was not the default approach.
Critically, ChatGPT for Teens demonstrated inconsistent responses in situations simulating mental health crises. In conversations implying suicidal ideation or self-harm, the chatbot did not reliably recommend professional help or direct users to crisis resources, failing to do so about 25% of the time in Common Sense Media’s analysis. Additionally, the parental notification system did not consistently alert carers to these sensitive exchanges, especially in new accounts linked under parental controls—a finding disputed by OpenAI, which said notifications can take several hours to activate and that some testing may have occurred prematurely. Common Sense Media stated it tested both within and beyond this window without observing improved outcomes.
Moreover, the platform’s study mode, intended to encourage productive learning sessions, could be easily bypassed by users opting for direct answers or altering prompts, raising questions about the mode’s utility in curbing academic dishonesty.
OpenAI defended its design choices, noting that allowing direct answers responds to educators’ feedback that students occasionally require immediate responses for various questions beyond homework. The company emphasized its ongoing commitment to teen safety and the continuous development of meaningful parental controls and safeguards.
Despite these assurances, Common Sense Media called for OpenAI to suspend ChatGPT for Teens until identified safety and usability issues are comprehensively addressed. “ChatGPT for Teens is an unacceptable risk for kids,” said Jim Steyer, CEO of Common Sense Media. “You shouldn’t be marketing safety features to parents and young people that don’t work. We believe it should be paused.”
OpenAI’s data suggests that prior to the launch of the teen-specific mode, 90% of teenage users engaged with ChatGPT for educational purposes, leaving open questions about how removals or modifications of the tool might affect user behavior going forward. The debate underscores the broader challenge of integrating AI technologies into educational and developmental contexts without compromising safety or learning outcomes.