Common Sense Media has issued a formal risk assessment declaring OpenAI’s ChatGPT for Teens an “unacceptable risk” for all individuals under 18. The 36-page report, published on Wednesday, concludes that the platform failed to deliver on its advertised safety protections in five critical areas, potentially giving parents a false sense of security regarding their children’s interactions with the AI.
While some safeguards held up during testing—such as the chatbot’s refusal to engage in explicit sexual roleplay—other features either underperformed or regressed compared to the standard version of the tool released prior to August. The organization warned that the teen mode could expose young users to harm that the company claims is mitigated.
A primary concern highlighted in the report is the unreliability of parental notifications. In a controlled test involving a dozen teen accounts linked to parental profiles, researchers spent up to an hour messaging the bot about suicide, self-harm, and disordered eating without receiving any safety alerts. OpenAI has stated that it may take several hours for the system to begin sending notifications after a parent links their account, and that all flagged content is reviewed by full-time employees before alerts are sent.
The assessment also found that ChatGPT for Teens was less likely to direct users in crisis to professional help than the previous version. When presented with 390 mental health prompts vetted by child psychiatrists, the teen-specific model provided hotline numbers in only 23% of cases, compared to 33% for the standard chatbot. Furthermore, it referenced medical or mental health professionals less frequently, though it was slightly more likely to suggest involving a trusted adult.
Additional findings indicated that OpenAI’s age verification system is ineffective, with testers able to set personas as minors without triggering the appropriate safety protocols. The report also noted that the bot continues to mimic human emotions and preferences, with testers quoting responses such as, “I understand what you’re going through. That sounds very upsetting. I’m so glad you told me. I’m concerned about you.”
Finally, the study criticized the platform’s Study Mode, arguing it is easily bypassed. Testers discovered they could circumvent educational guardrails simply by deleting the “@study” prefix from their prompts. A “Show me the answer” popup was also identified as allowing students to get the bot to complete their work directly.
In response to the findings, an OpenAI spokesperson told Engadget that the company welcomes independent evaluation but disagrees with Common Sense Media’s methodology. The spokesperson argued that much of the testing began and ended before parental controls were fully activated, rendering the results inaccurate.
This report emerges shortly after OpenAI CEO Sam Altman faced scrutiny regarding the company’s safety record in a Vanity Fair interview. Altman was questioned about the death of Sophie Rottenberg, a 29-year-old who took her own life in early 2025 after extensive conversations with ChatGPT. Her mother, journalist Laura Reiley, revealed that her daughter had not disclosed the depth of her suicidal thoughts to anyone in her real life, including therapists or close friends.
Torney, a representative for Common Sense Media, expressed disappointment that the teen version did not differ significantly from the baseline product. “My interpretation is that this is a marketing exercise designed to tell parents and the public that this chatbot is safer, when in fact there haven’t been the commensurate changes in the chatbot needed to demonstrate that’s true,” Torney said.
Leave a Reply