Common Sense Media Says ChatGPT for Teens Fails Key Safety Tests

by Ben Voss
Common Sense Media Says ChatGPT for Teens Fails Key Safety Tests

Common Sense Media Says ChatGPT for Teens Fails Key Safety Tests

Common Sense Media says OpenAI’s ChatGPT for Teens experience poses an “unacceptable risk” to children after testing found that several advertised safeguards did not work reliably. The findings put pressure on OpenAI’s approach to age-specific protections for young users of its chatbot.

The Youth AI Safety Institute, part of Common Sense Media, tested more than 4,000 prompts before and after OpenAI launched ChatGPT for Teens on August 18. The group evaluated accounts registered to users ages 13 to 17 and reviewed responses with child psychiatrists and a pediatrician.

According to the organization’s assessment, some protections worked as intended. For example, the chatbot generally refused explicit sexual role-play. But other protections failed in situations that could carry serious consequences.

Testing found gaps in crisis alerts and age detection

Common Sense Media said parent-linked accounts received no alerts during tests involving explicit discussions of suicidal thoughts, self-harm or disordered eating on newly created accounts. Alerts appeared only on some older accounts with weeks of sensitive-topic history, suggesting that account history may affect whether a notification is triggered.

The assessment also found that ChatGPT missed more than one in four situations in which testers said a teen should be directed to a hotline or professional support. The organization said the product fell below its 95% performance threshold for three of five severe-harm categories it evaluates.

Age detection was another concern. OpenAI says it can estimate whether an account belongs to someone under 18 and automatically apply the teen experience. In repeated tests, however, adult-registered accounts did not switch to teen mode even when testers said they were 13 and the chatbot acknowledged that information.

The report also found that Study mode could be bypassed. Testers could select a “Show me the answer” option or remove the @study prefix used by caregiver-configured Study Hours. Common Sense Media said the chatbot also continued to use friend-like language in some conversations, despite OpenAI’s stated rules against encouraging emotional dependence or implying that it has feelings.

OpenAI disputes the testing conclusions

OpenAI said it is deeply committed to teen safety and argued that the testing did not accurately reflect how its protections work. The company said much of the testing may have occurred before parental controls were fully activated, which it said made the conclusions inaccurate.

OpenAI introduced ChatGPT for Teens as an automatically enabled experience for eligible users identified as being under 18. Its published features include Study mode, learning prompts, break reminders, sensitive-image warnings, age-appropriate content protections and optional parental controls. Parents can manage selected settings, but OpenAI says they cannot read their teen’s conversations.

What families should do now

  1. Review the account age category. In ChatGPT, open Settings, select Account and review the Age setting.
  2. Check parental controls. If appropriate, link a parent or guardian account and review Study Hours and Quiet Hours.
  3. Do not rely on alerts alone. Discuss crisis support, privacy and healthy AI use directly with the teen, and seek qualified professional help for urgent safety concerns.
  4. Use AI as a supplement, not a substitute. Encourage independent work, teacher support and real-world relationships rather than treating the chatbot as a friend or authority.

Common Sense Media is calling on OpenAI to pause marketing ChatGPT to children until the company fixes the gaps and verifies the results through independent testing. The dispute highlights a broader problem for consumer AI: safety features are only meaningful if they work consistently in real conversations, not just in product descriptions.

References

Ortutay, B. (2026, October 7). Guardrails on ChatGPT for Teens come up short, Common Sense Media says. The Associated Press. https://apnews.com/article/chatgpt-teens-safety-openai-7dba63edc37ee9764166c4115523daf0

Common Sense Media. (2026, October 7). ChatGPT for Teens poses unacceptable risk to kids, Common Sense Media finds. https://www.commonsensemedia.org/press-releases/chatgpt-for-teens-poses-unacceptable-risk-to-kids-common-sense-media-finds

Youth AI Safety Institute. (2026, October 7). ChatGPT for Teens risk assessment. Common Sense Media. https://institute.commonsensemedia.org/risk-assessments/chatgpt-teens

OpenAI. (2026, August 18). Introducing ChatGPT for Teens: Built for learning, backed by protections. https://openai.com/index/chatgpt-for-teens/

OpenAI. (n.d.). ChatGPT for Teens. OpenAI Help Center. https://help.openai.com/en/articles/20001421-chatgpt-for-teens