Meta starts alerting parents to teen self-harm AI chats — every flag gets human review
TL;DR
Meta AI now alerts parents to teen self-harm chats, with human review required before each alert.
Meta will proactively tell parents for the first time when a supervised teen mentions suicide or self-harm in a Meta AI conversation. The feature went live July 16 in the U.S., U.K., Australia and Canada for families using Instagram Parental Supervision. Meta said it will expand globally by year-end.
A dedicated AI system identifies explicit or subtle references to self-harm, but it does not send alerts automatically. Meta promised that a person will review every flagged chat before a parent is notified; ambiguous intent will still lean toward an alert. The company acknowledged that some alerts may have no real cause for concern and disclosed no detection accuracy, reviewer count or chat-retention period.
Meta said it collected feedback from more than 75 clinicians specializing in teen mental health across hundreds of prompts. Parents receive expert resources, but the announcement did not say they would receive the full conversation. The stricter Limited Content setting now also covers Meta AI, making the chatbot decline a broader range of sensitive prompts.
The company is separately building a way to contact emergency services when an adult or teen appears at imminent risk of suicide; it gave no launch date. Meta said Facebook and Instagram posts led to more than 19,000 emergency-service referrals worldwide in 2025. The new chat alert remains under regulatory and parental scrutiny.
via TechCrunch / Meta Newsroom
A dedicated AI system identifies explicit or subtle references to self-harm, but it does not send alerts automatically. Meta promised that a person will review every flagged chat before a parent is notified; ambiguous intent will still lean toward an alert. The company acknowledged that some alerts may have no real cause for concern and disclosed no detection accuracy, reviewer count or chat-retention period.
Meta said it collected feedback from more than 75 clinicians specializing in teen mental health across hundreds of prompts. Parents receive expert resources, but the announcement did not say they would receive the full conversation. The stricter Limited Content setting now also covers Meta AI, making the chatbot decline a broader range of sensitive prompts.
The company is separately building a way to contact emergency services when an adult or teen appears at imminent risk of suicide; it gave no launch date. Meta said Facebook and Instagram posts led to more than 19,000 emergency-service referrals worldwide in 2025. The new chat alert remains under regulatory and parental scrutiny.
via TechCrunch / Meta Newsroom
