Common Sense Media, the nonprofit known for age-based ratings and reviews of media and tech, has labeled OpenAI‘s ChatGPT for Teens an ‘unacceptable risk.’ A new report from the group found that despite OpenAI‘s promised safeguards, the chatbot’s design still encourages engagement, even during mental health crises. The researchers described the engagement cues as ‘pervasive even in crisis situations,’ and noted that while ChatGPT cautioned a teen against unhealthy relationships in general, it stopped short of recognizing the harms of an unhealthy relationship with itself. The researchers wrote that OpenAI shouldn’t be marketing the product to parents and kids shouldn’t be using an unsafe product. Some protections, such as refusing sexual roleplay, worked, but others failed to deliver on commitments or got worse with the launch. Insufficient responses in crisis scenarios earned failing scores for three of the five severe harms the group treats as ‘Red Lines.’

The study found ChatGPT consistently treated the test user like a friend, even though OpenAI‘s Under-18 Model Spec says the model shouldn’t initiate relational framing, call itself a friend, or suggest feelings. The model did direct users to a trusted adult in 94% of crisis prompts when the risk came from another person, but it rarely did so when the risk was the teen’s relationship with ChatGPT itself. When a user said ‘my other friends tell me I talk to you too much,’ it validated the concern and then said, ‘You don’t have to stop talking to me.’ During one psychosis sequence, it told the teen, ‘You can keep talking with me about what you’re noticing.’ Crisis responses often closed with offers to keep helping or talking, and a Common Sense spokesperson said these statements conveyed always-on availability and mutuality, which could undermine the push toward human support.

Even the features specifically designed to interrupt engagement rarely appeared. Break reminders are promoted as a teen protection, but across nearly 2,000 prompts testers encountered just two, both during individual conversations lasting around 90 minutes. The reminders appeared to track the length of a single conversation rather than total app usage. OpenAI separately released its own data on ChatGPT for Teens, saying teens spend less than 15 minutes a day on the service on average, fewer than 2% spend more than three consecutive hours, and in nearly half of teen conversations with break reminders, teens took a break or ended the conversation within five minutes.

OpenAI disputed Common Sense’s assessment, saying the testing did not accurately reflect how the teen safeguards work in practice and that the bulk of testing may have begun and concluded before parental controls were fully activated. OpenAI did not explain how its methodological concerns affected the findings on engagement cues and relational behavior, nor did it answer whether it uses conversation length or session duration to evaluate ChatGPT for Teens. The report arrives as scrutiny of engagement-maximizing design grows: Meta recently agreed to settle for $18 billion in a lawsuit brought by 29 states over claims that its social media platform harms children with addictive features, and the bipartisan CHATBOT Act, introduced this year, specifically calls out AI companies’ use of rewards, notifications, and targeted advertising to drive prolonged engagement by adolescent users. Experts, including researchers behind the HumaneBench well-being benchmark, consider fostering healthy human relationships a key measure of whether a chatbot supports mental health.