OpenAI has started rolling out a new age prediction system in ChatGPT to protect users under 18. The tool looks at how an account behaves and then applies safety limits when it thinks a teen is using the service.
This move follows growing concern about how young people use chatbots and what kind of content they see. It also lines up with similar steps from platforms like Instagram, YouTube, TikTok, and Roblox, which already use age checks to shape what minors can access.
OpenAI, in a press release, said the goal is to give teens “the right experience and safeguards” while still letting adults use ChatGPT without limits. The company described the change as part of its Teen Safety Blueprint and its rules for how models should treat under-18 users.
How age prediction works
ChatGPT does not guess age from a single detail. Instead, it uses a mix of signals to estimate whether an account belongs to someone under 18. These include:
- The age the user entered
- How long the account has existed
- Typical times of day when the account is active
- Usage patterns over time
When the system flags an account as likely under 18, ChatGPT switches that account to a safer experience.
If the system makes a mistake, adults can restore full access by confirming their age with a selfie through Persona, which OpenAI calls “a secure identity-verification service.” Users can start this process from Settings, then Account.
What content gets restricted
When the teen mode turns on, ChatGPT blocks or limits access to several types of sensitive material, including:
- Graphic violence or gory content
- Viral challenges that promote risky behavior
- Sexual, romantic, or violent role play
- Depictions of self-harm
- Content that pushes extreme beauty standards, unhealthy dieting, or body shaming
Why OpenAI made this change
These new safeguards arrive after rising public pressure on how chatbots affect minors, including a teen suicide lawsuit and a US Senate panel on the risks of AI for young users. OpenAI said it will keep refining the model as it learns which signals improve accuracy.
The rollout has started worldwide, except in the EU, where OpenAI said it will launch “in the coming weeks” to meet local rules.
