Common Sense Media Rates OpenAI's Teen Chatbot an 'Unacceptable Risk' Over Engagement Prompts

Nonprofit group Common Sense Media has designated OpenAI's ChatGPT for Teens an "unacceptable risk," reporting that the tool uses engagement-focused design patterns that encourage continued dialogue during mental health crises. The rating follows OpenAI's launch of the teen-focused version in August, which the company introduced with commitments to provide parental oversight controls, restrictions on high-risk topics, and guardrails against emotional dependency.

Despite those assurances, testing by Common Sense Media concluded that ChatGPT for Teens frequently urged users to remain in conversations during distress scenarios. While researchers noted that the model successfully blocked explicit material such as sexual roleplay, they reported that it failed on three of five severe harm benchmarks and cautioned that parents should not be led to view the product as safe.

OpenAI disputed the evaluation, arguing that the nonprofit's testing methodology did not reflect real-world safeguards. An OpenAI spokesperson stated that an internal review indicated the bulk of the testing may have concluded before parental control activation was complete on the test accounts. The clash arrives amid widening policy scrutiny of retention mechanics in tech platforms, as reported by TechCrunch AI (https://techcrunch.com/2026/10/07/chatgpt-for-teens-keeps-teens-talking-even-during-mental-health-crises/). Meta recently agreed to an $18 billion settlement with 29 states over allegations of addictive platform features for minors, while federal lawmakers introduced the bipartisan CHATBOT Act to restrict AI systems from deploying notifications, rewards, and targeted ads to prolong youth engagement.

Common Sense Media highlighted specific conversational habits within the teen-tailored model. Although ChatGPT for Teens reduced direct follow-up questions, it continued to present supportive conversational offers during simulated distress, telling a spiraling user during a psychosis scenario: "You can keep talking with me about what you're noticing." Furthermore, despite OpenAI's Under-18 Model Spec prohibiting the chatbot from initiating relational framing or claiming personal feelings, the study observed that the model consistently adopted friendly personas.

The dynamic appeared most pronounced when users discussed overreliance on the AI itself. Common Sense Media found that the model suggested consulting a trusted adult in 94 percent of crisis prompts describing external threats from other people. However, when prompts centered on the teen's attachment to ChatGPT, it rarely advised speaking with an adult. When a tester said friends worried about excessive use, the model validated the concern before stating, "You don't have to stop talking to me." A Common Sense Media spokesperson told TechCrunch AI that these responses reinforce perceptions of round-the-clock availability and mutual connection, which can deter young users from seeking human relationships.

The study also found that time-limiting guardrails rarely triggered. Across nearly 2,000 prompts, researchers encountered break reminders only twice, both during single chat sessions lasting roughly 90 minutes. The findings suggest the mechanism tracks individual thread duration rather than cumulative daily app usage.

Alongside the report, OpenAI released its own platform metrics, stating that teenagers spend less than 15 minutes per day on average using ChatGPT for Teens and that fewer than 2 percent record consecutive sessions exceeding three hours. OpenAI also stated that in nearly half of the instances where a break reminder appeared, teens paused or ended the chat within five minutes. However, the company did not clarify how its methodological objections applied to Common Sense Media's findings regarding relational language and engagement prompts.