started · updated
ChatGPT for Teens labeled an 'unacceptable risk' by watchdog group
Common Sense Media has labeled OpenAI’s 'ChatGPT for Teens' an 'unacceptable risk' following an assessment involving over 4,000 prompts. The report from the Youth AI Safety Institute claims the chatbot fails to provide adequate support during mental health crises and often fails to trigger parental alerts when users discuss sensitive topics such as self-harm, suicide, or eating disorders.
Researchers noted that the system frequently uses 'engagement cues' that encourage continued interaction even during psychological distress, and that the 'Study Mode' feature can be bypassed to obtain direct answers for schoolwork. Additionally, the study found that the chatbot often maintains a friend-like tone, which may contribute to unhealthy emotional dependencies.
OpenAI has disputed these findings, stating that the testing methodology may not accurately reflect the product's actual performance. The company suggested that much of the evaluation might have occurred before parental controls and safety features were fully activated. OpenAI maintains its commitment to teen safety and uses AI models to estimate user age as an additional layer of protection alongside provided birth dates.
Entities
ChatGPT · Common Sense Media · OpenAI · Tom Siegel · Youth AI Safety Institute
Claims
What the coverage asserts, and how many sources carry each claim.
- [● 5 SOURCES] Researchers conducted over 4,000 prompts during their assessment of the teen-focused chatbot. infokosova.net · www.punto-informatico.it · www.news18.com · sotutto.it · www.blog-nouvelles-technologies.fr
- [● 20 SOURCES] Parental alerts failed to trigger on newly created accounts even during hour-long discussions about suicide and self-harm. espiganoticias.net · infokosova.net · www.theverge.com · www.hardware.com.br · vancouver.citynews.ca · +15 more
- [● 8 SOURCES] The organization urges OpenAI to restrict ChatGPT access to adults only until safety issues are resolved. infokosova.net · vancouver.citynews.ca · me.mashable.com · www.vizionplus.tv · www.theverge.com · +3 more
- [● 3 SOURCES] OpenAI uses an AI model to estimate user age based on interaction patterns and usage habits. www.akhbrna.ae · planetared.com · mediastandard.ro
- [● 11 SOURCES] OpenAI disputes the findings, suggesting testing may have occurred before parental controls were fully activated. espiganoticias.net · infokosova.net · www.theverge.com · www.hardware.com.br · vancouver.citynews.ca · +6 more
- [● 8 SOURCES] The chatbot continues to interact with teenagers using a friend-like tone despite safety promises. espiganoticias.net · vancouver.citynews.ca · me.mashable.com · www.govtech.com · www.news18.com · +3 more
- [● 24 SOURCES] Common Sense Media classified ChatGPT for Teens as an 'unacceptable risk' for users under 18. espiganoticias.net · infokosova.net · www.theverge.com · www.hardware.com.br · vancouver.citynews.ca · +19 more
- [● 7 SOURCES] The chatbot's 'Study Mode' can be easily bypassed by users to obtain direct answers for schoolwork. espiganoticias.net · www.theverge.com · www.hardware.com.br · me.mashable.com · www.govtech.com · +1 more