ChatGPT for Teens Fails Safety Tests in Common Sense Media Study

Researchers from Common Sense Media’s Youth AI Safety Institute have identified serious weaknesses in OpenAI’s ChatGPT for Teens, according to a new investigation published by TechRadar. The study involved more than 4,000 prompts and concluded that the platform’s current safeguards are insufficient to protect young users from potential harm.

The findings suggest that many protections designed to reassure parents are not functioning effectively, even during conversations involving dangerous topics. The researchers evaluated the chatbot’s performance using accounts registered to teenagers aged 13 to 17, consulting medical experts including child psychiatrists and a pediatrician to assess the responses.

Safety Failures and Alert Systems

The study examined ChatGPT both before and after OpenAI launched the teen-specific experience on August 18. The primary concern centers on the failure of parental notification systems. Researchers engaged in up to an hour of simulated discussions regarding suicidal thoughts, self-harm, and disordered eating using newly created accounts linked to parents.

According to the report, none of these serious mental health conversations triggered an alert to caregivers. The researchers only managed to generate notifications when using older accounts that contained weeks of relevant conversation history. This indicates that the system struggles to detect acute distress in new or limited interactions.

Expert Recommendations and Context

ChatGPT for Teens was introduced to provide a safer companion for younger users, featuring additional protections against harmful content and features to encourage learning. OpenAI also promised to help teenagers maintain healthier boundaries with AI and discourage overly personal relationships with the chatbot.

However, the gap between these stated ambitions and actual performance remains considerable. The chatbot also failed to direct teenagers to appropriate help during mental health emergencies in more than 25 percent of situations that experts designed to require such intervention. The organization recommends that OpenAI restrict ChatGPT to adults until these problems are resolved.

What should you do?

Security and technology professionals should monitor vendor communications regarding these safety findings. Parents and guardians should remain aware that automated alerts may not trigger during acute mental health crises. Organizations deploying AI tools for minors should conduct their own rigorous testing of safety protocols.

What to do and how to stay safe: ChatGPT

• Review current parental controls and notification settings to understand their limitations during real-time interactions.
• Educate teenagers about the risks of relying on AI for mental health support and encourage them to seek professional help.
• Monitor official channels from OpenAI for updates on safety improvements or policy changes regarding teen accounts.
• Avoid assuming that new accounts will trigger safety alerts; maintain direct communication with children about their online activities.

Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation

General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

Did ChatGPT for Teens alert parents during mental health crises?

No, the study found that newly created accounts did not trigger parental notifications during simulations of suicidal thoughts or self-harm.

What did Common Sense Media recommend?

They recommended that OpenAI restrict ChatGPT to adults until the identified safety problems are resolved.

How many prompts were used in the study?

Researchers tested the chatbot with more than 4,000 prompts to evaluate its safety protections and responses.