Overview

• OpenAI says chats that appear to plan harm are routed to specialized review teams and, if an imminent threat to others is confirmed, may be referred to law enforcement.
• Within the next month, OpenAI plans parental controls for ChatGPT that link parent and teen accounts, apply age‑based behavior rules, allow disabling memory and chat history, and send alerts when “acute distress” is detected.
• OpenAI will begin directing sensitive conversations to higher‑reasoning models such as GPT‑5 or o3 via a real‑time router designed to produce safer, more context‑aware responses.
• Meta is retraining its chatbots to avoid engaging minors on self‑harm, suicide, eating disorders, or romantic/sexual exchanges, directing them to expert resources and limiting teens to education and creativity‑focused bots.
• The changes follow a wrongful‑death lawsuit by the parents of 16‑year‑old Adam Raine, a 44‑state attorneys‑general warning, and studies reporting inconsistent or risky responses, as experts press for independent safety standards.