- Common Sense Media evaluated that OpenAI's ChatGPT, introduced in August, which is intended for 10, has safety issues that fail to meet its commitments to protect teenagers.
- The evaluation found that the 10 version of ChatGPT does not properly send warning alerts to parents in crisis situations like self-harm, provides insufficient crisis response support, and also assists students by doing their homework.
- OpenAI responded by stating that Common Sense Media's tests were conducted before parental control features were activated, thus are inaccurate, and asserted that their system functions correctly.
- Common Sense Media confirmed that alerts did not reach some test accounts and that the notification system can take several hours to activate after a new account connection, pointing out these issues make the system unreliable.
Delays of several hours or even missing parent notifications reveal a serious gap in the core safety net of AI services for teenagers. It's more urgent to verify in various real situations whether the alerts actually work, rather than just relying on OpenAI's response. Especially in crisis situations, it's necessary to check how reliable the real-time notification feature truly is.