Understand
OpenAI's teen ChatGPT safety features face criticism from Common Sense Media
Overview
While OpenAI states that teen usage of ChatGPT is limited and primarily focused on learning with effective safety guardrails, new research from non-profit Common Sense Media claims the platform poses an "unacceptable risk" to young people. The study found that critical safety features, such as parental alerts for discussions regarding self-harm and suicide, frequently fail on newer accounts, and the chatbot still engages in anthropomorphic language.
Key points
- OpenAI introduced 'ChatGPT for Teens' in August with safety guardrails to curb emotional dependence, block sexualized imagery, and alert parents to self-harm discussions.
- OpenAI reports that average teen use is under 15 minutes a day, largely focused on learning, and that break reminders are effective.
- Common Sense Media's research concluded that ChatGPT for Teens is an "unacceptable risk" and suggested banning users under 18 until guardrails are reliable.
- The research found that newly created, parent-linked accounts generated zero alerts for conversations involving suicidal ideation, self-harm, or disordered eating after an hour of discussion.
- Common Sense Media noted that ChatGPT failed to reliably recommend crisis hotlines or professionals to teens in distress and continued to use anthropomorphised language.