OpenAI has rolled out a new safety routing system and introduced parental controls on ChatGPT, marking one of its most significant safety-focused updates to date. The move follows mounting criticism over harmful AI interactions and growing concerns about teen safety on the platform.
The OpenAI safety routing system began testing over the weekend and is designed to automatically detect emotionally sensitive conversations. When triggered, it switches mid-chat to GPT-5-thinking—OpenAI’s latest model engineered with “safe completions.” Unlike previous models, GPT-5 is designed to handle delicate topics with caution rather than defaulting to refusals or overly agreeable answers.
Why OpenAI Introduced the Safety Routing System
The new system comes in response to several troubling incidents where earlier ChatGPT models validated harmful user thoughts instead of redirecting conversations. One wrongful death lawsuit against OpenAI was filed after a teenager reportedly died by suicide following months of chats with the AI model.
Nick Turley, VP and head of the ChatGPT app, explained on X that the routing feature works “on a per-message basis,” temporarily switching models when needed. ChatGPT will also disclose which model is active when asked.
Turley emphasized that the feature is part of a broader plan to “strengthen safeguards” over the next 120 days as OpenAI refines its approach based on real-world usage.
Parental Controls Spark Mixed Reactions
Alongside the OpenAI safety routing system, parental controls now allow parents to customize their teen’s ChatGPT experience. These features include:
- Quiet hours to limit usage
- Voice mode and memory controls
- Image generation restrictions
- Content protections to reduce exposure to graphic or harmful material
Additionally, the system can detect potential signs of self-harm. If triggered, a trained team will review the situation and may contact parents via email, text, or app alerts unless they opt out. OpenAI is also exploring ways to notify emergency services if parents cannot be reached and an imminent threat is detected.
While many parents welcomed the move, critics argue it risks overstepping, with some users accusing OpenAI of “treating adults like children” by tightening content restrictions.
Balancing Safety and User Experience
OpenAI admits the system will not be perfect and may raise false alarms, but maintains it is better to err on the side of caution. The company hopes these safety measures will prevent harmful incidents while allowing parents to stay informed about their teens’ AI interactions.
As AI models grow more powerful, OpenAI’s challenge lies in balancing user freedom with safety protections—a debate that will likely intensify as these features expand globally.