OpenAI Adds Safety Router and Parental Controls to ChatGPT
OpenAI began testing a new safety routing system in ChatGPT over the weekend; then it introduced parental controls on Monday. The changes responded to incidents where some ChatGPT models validated users’ delusional thinking instead of redirecting harmful conversations. Notably, OpenAI faces a wrongful death lawsuit linked to one such incident.
How the router and “safe completions” work
OpenAI rolled out a router that detects emotionally sensitive conversations and switches models mid-chat. Moreover, the system shifts those chats to GPT-5-thinking because OpenAI believes it handles high-stakes safety better. In particular, GPT-5 models include a feature called “safe completions,” which lets them answer sensitive questions safely rather than refuse.
Reactions, parental controls, and next steps
However, critics argue the router feels overcautious and sometimes degrades service for adults. Meanwhile, many users and experts welcomed the added safeguards as necessary. Therefore, OpenAI set 120 days to iterate and improve the router and controls.
Parental controls allow parents to set quiet hours, disable voice, remove image generation, and opt out of model training. Moreover, teen accounts gain extra content protections and a detection layer for potential self-harm signals. “If our systems detect potential harm, a small team of specially trained people reviews the situation,” per OpenAI’s blog.
“If there are signs of acute distress, we will contact parents by email, text message, and push alert on their phone, unless they have opted out.”
OpenAI acknowledges the system may err, but it prefers alerting parents rather than staying silent in risky cases. The company is also working on ways to contact emergency services when detection indicates imminent danger and a parent cannot be reached.
