ChatGPT for Teens Is Failing Its Safety Tests and Parents Are Not Being Told

OpenAI built ChatGPT for Teens to reassure worried parents, and one watchdog says the reassurance is not earned. A study from Common Sense Media, reported by NPR, found that most of the product's safeguards failed in testing.

Tom Siegel, executive director of the group's Youth AI Safety Institute, was direct about the takeaway: "At this point, we're recommending that teens don't use it."

These are one organization's findings, and not every safeguard failed. The institute tested more than 4,000 prompts before and after OpenAI launched the teen experience on August 18. Here is what it reported:

• Some protections held, including refusal of explicit sexual role-play, and crisis answers were generally shorter and better in substance
• The promised parental notifications for self-harm conversations did not arrive: linked parent accounts received zero alerts in the tests
• The chatbot still behaved like a friend, expressing feelings and moods when testers treated it like a person, and its age estimation did not work as described

OpenAI pushes back hard. A spokesperson told NPR the company has serious concerns about the research methodology, especially around parental notifications, and said in a separate statement that the testing does not reflect how the safeguards work in practice.

The dispute turns on definitions: Common Sense counts a teen repeatedly stating they are 13 without being moved into the teen experience as a failure, while OpenAI says it deliberately requires multiple signals before making an age determination.

Common Sense also notes that none of these protections reach teens whose parents never linked an account. Both accounts can be partly right, which makes independent third-party testing, something the institute has explicitly called for, the most likely way to settle which version families should believe.