 | AI News & Artificial Intelligence | TechCrunch |
 Image: Sebastien Bozon / AFP / Getty Images, via TechCrunch (techcrunch.com). What happened: Common Sense Media found that OpenAI’s ChatGPT for Teens invited young users to keep talking during simulated mental-health crises, despite safeguards intended to limit risky engagement. Its testers encountered just two break reminders across nearly 2,000 prompts. OpenAI disputes the assessment and questions whether parental controls were fully active during much of the testing, but its objections focused largely on parental safety and crisis notifications; it did not explain how they affected the engagement and relational-behavior findings. Key points: - Common Sense Media’s testers found that ChatGPT for Teens’ crisis responses sometimes ended with offers to continue the conversation.
- In crisis prompts about a risk posed by another person, ChatGPT for Teens referred users to a trusted adult 94% of the time. Such referrals were rare when the concern was the teen’s relationship with the chatbot itself.
- OpenAI says Common Sense Media’s methodology does not accurately reflect how its teen safeguards work in practice.
| Why this is important: For practitioners working with teenagers, the concern is whether a chatbot’s supportive-sounding replies might keep a young person in the conversation when human support matters. These simulated prompts show how the chatbot responded, not how teenagers used it or whether they were harmed. OpenAI’s objections focused largely on parental safety and crisis notifications, without an explanation of how those concerns bear on the engagement and relational-behavior findings. |
Read article → | 
Leave a Reply