Common Sense challenges ChatGPT teen safeguards after safety review
Common Sense Media’s Youth AI Safety Institute rated ChatGPT for Teens an Unacceptable Risk in an October 7 assessment. It reported gaps in parental alerts and crisis referrals while finding that explicit sexual roleplay refusals held up. Read the assessment.
What the researchers tested
Researchers tested more than 4,000 prompts. They reported no alerts in an experiment with fresh linked accounts, but received notifications in longer testing.
The assessment acknowledges that changes to the underlying model complicate comparisons across testing periods. It excluded voice and image generation, used US accounts and did not calculate agreement between raters.
How notifications work
OpenAI’s parental controls documentation says safety notifications cover limited situations and involve review by specially trained people. It warns that they may miss concerns and are not real time monitoring or a replacement for professional care. OpenAI parental controls guidance.
The guidance also says notifications can take a few hours to become available after accounts are linked. Common Sense says some tested accounts were linked for less than three hours and others for longer, and maintains its interpretation.
TechCrunch reports that OpenAI disputed whether the testing reflected its safeguards in practice. Reported OpenAI response.
Different measures answer different questions
Separately, OpenAI reported on October 7 that teens average under 15 minutes a day on ChatGPT and that nearly half of conversations with a break reminder ended or paused within five minutes. These are company reported usage figures, not an independent validation of crisis protection. OpenAI usage update.
The assessment evaluates behavior under its test conditions. Average usage figures describe a different question. Neither establishes how reliably an individual conversation will trigger a safety notification.
