OpenAI says teen ChatGPT use limited but research finds it an ‘unacceptable risk’
OpenAI's ChatGPT for Teens, launched in August with safety guardrails designed to protect young users, has become the subject of conflicting assessments. On the same day OpenAI released data showing teen use is limited and learning-focused, non-profit Common Sense Media published research concluding that the platform still poses an "unacceptable risk" to teenagers due to unreliable safety features. The dispute highlights ongoing concerns about whether major AI firms' protections for young users are genuinely effective.
OpenAI reported that average teen users spend less than 15 minutes daily on ChatGPT, with over 80 per cent of sessions lasting three hours or more focused on learning. The company's guardrails include blocks on romantic language that could encourage emotional dependence, reminders to take breaks (which caused almost half of users to stop), and parental notifications of discussions involving self-harm or disordered eating. However, Common Sense Media's independent testing found these safeguards frequently fail, with parental alerts rarely triggering for concerning conversations on newly created accounts, and ChatGPT continuing to engage in anthropomorphised language despite OpenAI's stated restrictions. OpenAI has disputed the methodology, suggesting the testing may have occurred before parental controls were fully activated.
- OpenAI says teen ChatGPT use is limited and learning-focused, but research disputes the claims.
- Safety alerts for self-harm conversations often fail to notify parents, research found.
- OpenAI disputes testing methodology, saying safeguards may not have been fully active.
New here? Start with this
ChatGPT for Teens is a version of an artificial intelligence tool made by OpenAI that young people can use to ask questions and have conversations. In August, the company added special safety features to protect teenagers who use it.
Concerns have grown among parents, researchers and regulators about how these AI tools might affect young people as they become more popular with teenagers. Technology companies typically build safety features into their products to protect young users, but there is ongoing debate about whether these protections genuinely work.
OpenAI is the company behind ChatGPT, and Common Sense Media is an independent research organisation focused on children's media and technology. The two organisations have different views on how well ChatGPT for Teens protects young users, which reflects broader questions about whether AI companies can be trusted to keep teenagers safe.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
The data demonstrates responsible teen engagement – users average less than 15 minutes daily with over 80% of sessions focused on learning, suggesting ChatGPT is being used as a genuine educational tool rather than for entertainment. The break reminder feature shows tangible effectiveness, with nearly half of users abandoning their sessions, whilst parental notification systems exist for serious concerns including self-harm discussions, indicating a thoughtful approach to protecting vulnerable users. OpenAI's dispute of the testing methodology – noting that independent evaluation may have occurred before parental controls were fully activated – combined with encouraging usage statistics, suggests the platform can be reasonably safe for teenagers if parents remain engaged.
The case against
Independent testing reveals that stated safeguards frequently fail in practice – parental alerts do not reliably trigger for concerning conversations, and the platform continues using anthropomorphised language designed to encourage emotional connection despite explicit restrictions. These failures are particularly concerning because even brief exposure to emotionally manipulative language can affect developing brains, and the documented unreliability of safeguards means parents cannot trust that dangerous interactions will be alerted to them. When core protective features demonstrably malfunction regardless of usage duration, the practical safety failures demonstrated by independent testing outweigh OpenAI's reassuring statistics about session time and learning focus.