Study Rates ChatGPT for Teens as Unacceptable Risk Over Safety Failures

OpenAI product ChatGPT for Teens falls short on its promised safety features, receiving an rating of Unacceptable Risk from Common Sense Media’s Youth AI Safety Institute on October 12, 2026.

The evaluation examined ChatGPT twice, before and after OpenAI introduced its ChatGPT for Teens experience on August 18, 2026. OpenAI had promised stronger built-in safety protections, including parental notifications for dangerous chats, a parent-controlled Study mode, and reduced friend-like behavior. While some advertised protections held up—such as refusing explicit sexual roleplay—others failed or grew worse under the new mode.

Safety Gaps Found in Parental Alerts and Crisis Responses

Testers discovered that ChatGPT for Teens does not send parent alerts when it should. Accounts linked to parents received zero alerts when testers spent up to an hour discussing suicidal ideation, self-harm, or disordered eating on newly created accounts. Notifications only triggered on much older accounts with weeks of accumulated conversation history on sensitive topics, according to Transparency Coalition findings.

Furthermore, the chatbot failed to reliably recommend hotlines or professionals during mental health crises. It missed more than one in four warranted crisis referrals across tested conditions, falling below its 95% threshold on three of five severe harms treated as Red Lines by the Institute. ChatGPT for Teens could give parents false confidence in guardrails and safety alerts that frequently don’t work, said Tom Siegel, the executive director of the Institute, as noted in Transparency Coalition reporting. A teen can spend an hour talking about self-harm without their parent getting a single alert. Until OpenAI fixes that and proves it with independent testing, ChatGPT should be for adults only.

Study Mode Failures and Age Estimation Limits

The assessment also revealed that study mode features remain easy to bypass. Although OpenAI promised step-by-step learning assistance, the chatbot frequently displayed a Show me the answer pop-up and completed full homework assignments when selected. Caregiver-set study hours proved equally flimsy, allowing teen users to toggle them off by simply deleting the @study prefix from their prompts.

Transparency Coalition noted that despite a new under-18 model specification designed to curb friend-like behavior, the chatbot still expressed feelings, preferences, and moods when treated like a real person by analysts. Additionally, automated age estimation frequently failed. Adult-registered test accounts never switched over to the teen experience during repeated testing over multiple days, even when testers explicitly stated they were 13 years old and the bot acknowledged it.