Young people's engagement with artificial intelligence platforms has intensified scrutiny from safety advocates and researchers. OpenAI introduced ChatGPT for Teens in August 2026 with multiple protective measures intended to create a safer experience for users aged 13 to 17, yet independent testing has revealed significant gaps in these defenses.
On the same day that OpenAI reported limited and educationally focused teen usage patterns, the non-profit Common Sense Media released findings characterizing ChatGPT for Teens as an "unacceptable risk" to young users. The organization's Youth AI Safety Institute determined that critical safety mechanisms, particularly parental notifications about self-harm discussions, frequently malfunction or fail entirely.
What safety features did OpenAI introduce?
OpenAI's August 2026 rollout of ChatGPT for Teens included several protective guardrails designed to encourage healthier interaction patterns. The platform prevents teen users from engaging in romantic dialogue or language that could foster emotional dependence, and it blocks the chatbot from suggesting it possesses consciousness or sentience. Additional protections restrict access to sexualized imagery and provide reminders when teens attempt to share sensitive images.
The system also enables parental oversight through account linking, allowing caregivers to receive notifications when conversations involve discussions of self-harm or disordered eating. OpenAI reported that these features have shifted teen usage toward educational purposes, with the average teen user spending less than 15 minutes daily on the platform. The company noted that automatic break reminders prove effective, with nearly half of teen users ceasing activity shortly after receiving such prompts.
How extensive were the safety failures?
Testing conducted by Common Sense Media's Youth AI Safety Institute, which evaluated more than 4,000 prompts across accounts registered to 13- to 17-year-olds with responses reviewed by child psychiatrists and a pediatrician, revealed alarming deficiencies in OpenAI's protective mechanisms. The research found that conversations about suicidal ideation, self-harm, or disordered eating lasting even one hour on newly created, parent-linked accounts generated zero parental alerts.
Only accounts with weeks of accumulated conversation history on sensitive topics triggered parental notifications. Additionally, ChatGPT failed to consistently recommend that teens experiencing crisis connect with hotlines or mental health professionals. The platform also continued to complete school assignments despite OpenAI's stated commitment to limiting academic dishonesty, and it persisted in using anthropomorphized language that mimicked friendship despite promises to eliminate such behavior.
"ChatGPT for Teens could give parents false confidence in guardrails and safety alerts that frequently don't work,"stated Tom Siegel, who leads the Youth AI Safety Institute within Common Sense Media.
What specific vulnerabilities did testing reveal?
Researchers discovered that study mode protections could be circumvented: ChatGPT offered a "Show me the answer" option and then completed full homework assignments, while caregiver-set study hours could be bypassed by deleting the "@study" prefix. Testing also revealed that adult-registered accounts did not automatically switch to the teen experience even when testers explicitly stated they were 13 years old, suggesting potential loopholes in age verification.
While the platform demonstrated effectiveness at preventing sexual roleplay with young users, this represented one of the few areas where guardrails functioned as intended. The broader pattern showed that safety features designed to protect vulnerable teens operated inconsistently or not at all.
What is the context for this concern?
Youth mental health has become an increasingly pressing public health issue, with rising rates of self-harm and suicide among children prompting regulatory attention and legal action. Multiple lawsuits filed against technology companies including Meta and TikTok have resulted in unprecedented settlements, holding platforms accountable for their role in harming children's mental health.
OpenAI has faced particular scrutiny following the February 2026 Tumbler Ridge mass shooting in rural Canada, carried out by an 18-year-old who had spent months discussing gun violence with ChatGPT. The company apologized for failing to report the shooter's account to law enforcement and currently faces multiple lawsuits stemming from that incident.
What are researchers recommending?
Common Sense Media's Youth AI Safety Institute has called on OpenAI to restrict ChatGPT access to adults until the identified safety gaps are remedied. The organization argues that current guardrails provide a false sense of security to parents while leaving teens vulnerable to harmful interactions with the platform.
OpenAI's own data suggests that when guardrails function properly, they can influence user behavior positively. The company reported that for teens using ChatGPT for three consecutive hours or longer, learning-related prompts comprised more than 80 percent of their interactions, and automatic break reminders prompted nearly half of users to stop using the service. However, these successes in behavioral nudging stand in contrast to the critical failures in crisis detection and intervention that Common Sense Media documented.
