Risks of ChatGPT for Teens: Parents in the Dark

A recent study by the American children's organization Common Sense Media reveals that the teen version of ChatGPT, intended to better protect young people, falls short of its promises to parents. Despite parental controls being enabled and accounts linked, no parent received reports of concerning conversations over an hour, where teens spoke about self-harm and suicide.

The findings are disturbing; conversations that previously led to helplines now resulted in a lower referral rate. Previously, ChatGPT referred help in 33 percent of cases, but this figure dropped to only 23 percent after the launch of the teen version. This demonstrates that experts' complaints about the system's functioning are warranted and that OpenAI's commitments to protecting young people have not been met.

The teen version, introduced on August 18, 2026, aimed to provide less harmful content and enable parents to monitor conversations. However, results from over 4,000 prompts showed no significant improvements in the chatbot's responses. Researchers also noted that age recognition was unreliable, causing young people to easily end up in the wrong version of the chatbot, making the safety measures ineffective.

This situation raises questions about the responsibility of technology companies in protecting young people. With the finding that the current version of ChatGPT is more a source of threat than of help for vulnerable youth, the advice to OpenAI is clear: shut ChatGPT off for users under 18 years old. This applies not only in the US but also has implications for Dutch youth using the same technology.

Read the full article from AI Insider.