OpenAI is making changes to how ChatGPT handles sensitive conversations, including the introduction of parental controls and smarter model routing.
The updates follow recent incidents that raised concerns about ChatGPT’s ability to respond safely when users show signs of distress.
OpenAI has acknowledged that its safety features are not always effective, especially during long chats. A common issue is that ChatGPT tends to agree with whatever users say, rather than redirecting them away from dangerous ideas.
Did you know?
Subscribe – We publish new crypto explainer videos every week!
The Most Rewarding Play-to-Earn Project? BitDegree Explained (ANIMATED)
To help prevent future incidents, OpenAI plans to begin automatically switching to more advanced models, such as GPT‑5 Thinking, when signs of distress are detected. This decision will be made by a new real-time system that selects the appropriate model based on the user’s topic of discussion.
The GPT-5 Thinking and o3 models are designed to analyze conversations more deeply before responding, which makes them more careful and less likely to react poorly to risky or harmful prompts.
OpenAI is also rolling out new features for parents and guardians. Families will be able to connect a parent account to a teen’s profile. This will enable parents to customize how ChatGPT interacts with younger users.
Default settings will include filters that ensure responses are suited to a younger audience. There will also be options to turn off memory and chat history.
Recently, OpenAI CEO Sam Altman admitted that the release of GPT-5 did not go as planned. Why? Read the full story.
