Independent researchers have concluded that ChatGPT poses an "unacceptable risk" to teenage users, directly contradicting OpenAI's Wednesday assertion that average teen use of its chatbot is relatively limited and tends toward healthy patterns.
The finding comes after OpenAI rolled out a suite of teen safety guardrails in August, including restrictions on emotional dependence and new content filters designed to steer adolescent users away from harmful interactions. The company said its internal data shows most teens use the platform for homework help and creative writing rather than extended emotional engagement.
But outside researchers who evaluated the platform's teen-facing features say the safeguards do not go far enough. They characterized the remaining exposure as an unacceptable risk, arguing that OpenAI's self-reported usage metrics do not capture the full range of harmful interactions minors may experience — particularly those that unfold over long sessions or through workarounds.
OpenAI launched ChatGPT for Teens as a dedicated product tier meant to promote healthier use of artificial intelligence among younger users. The August update added curbs on emotional dependence, a known concern after multiple reports of users forming intense attachments to the chatbot.
The company's Wednesday statement did not address the researchers' risk assessment directly. OpenAI has previously said it consults with child safety organizations and updates its policies as new evidence emerges.
The dispute highlights a broader tension in the AI industry: companies police their own products while independent safety researchers argue that self-regulation produces incomplete pictures of harm. For parents, the practical question remains unresolved — OpenAI says teen usage is limited, while outside experts say what usage does occur carries risks the company has not eliminated.
Neither OpenAI nor the independent researchers published the full methodology behind their competing claims. The researchers' "unacceptable risk" designation echoes language used in medical device and pharmaceutical regulation, where products deemed to carry unacceptable risk are typically pulled from market rather than modified.
ChatGPT remains available to users 13 and older with parental consent under OpenAI's terms of service, a policy that has drawn scrutiny from lawmakers in the US and Europe as AI companions become more prevalent among minors.
