OpenAI says ChatGPT for Teens will make AI safer for young users. The launch comes amid lawsuits over whether ChatGPT is safe for children. But child safety experts say the company still has to prove its protections actually work. Here’s what you should know.
Launched in August 2026, ChatGPT for Teens is an age-gated experience that automatically becomes the default when a user is identified as under 18. Once activated, it applies stricter content rules and, in some cases, alerts parents with a linked account.
Child safety experts, however, are skeptical. Robbie Torney from Common Sense Media told Engadget(nytt fönster) that OpenAI makes important safety commitments, but it needs evidence that they actually work.
Josh Golin of Fairplay said safeguards may look promising on paper but lack accountability, and risk being more of a public relations response than a carefully tested safety system. Both questioned the very system that estimates age, while Golin also warned that ChatGPT’s existing restrictions on suicide, violence, and drug use have failed or weakened over time.
How ChatGPT for Teens works
ChatGPT estimates whether an account likely belongs to someone under 18 by combining account and behavioral signals, including:
- The age provided by the user
- How long the account has existed
- The general topics discussed
- Typical times of activity
- How the account’s usage develops over time
When the model predicts that the account may belong to a minor, ChatGPT automatically switches that account into the ChatGPT for Teens experience, where stricter safeguards apply, including limits on romantic language, emotional dependence, self-harm content, and conversations about eating disorders.
For teens whose accounts are linked to a parent’s, OpenAI may also send an alert when a conversation suggests a risk of self-harm. The company told Engadget that every flagged conversation would be reviewed by a full-time OpenAI employee before an alert is sent, with a goal of notifying parents within one hour of the prompt. That is an exceptionally ambitious commitment that child safety experts doubt OpenAI can pull off at scale.
Adults who are incorrectly placed in the teen experience can verify their age through Persona, a third-party age verification service. Depending on the country, Persona may request a live selfie, a government-issued ID, or both.
Age verification is coming for AI
Stronger guardrails can be a good thing, provided they work. But deciding who receives them comes with a privacy cost. This is AI surveillance: ChatGPT classifies people by analyzing what they discuss, when they use the service and how they use it over time.
Adults incorrectly classified as minors may have to verify their age through Persona. OpenAI says it does not receive the selfie or ID itself, and that Persona deletes the uploaded material within seven days. Even so, restoring full access may require users to hand sensitive identity information to a third-party verification service.
That privacy trade-off is harder to justify when the underlying protections remain unproven. As child-safety experts have warned, OpenAI has not yet shown that its age-prediction system, content safeguards or parental alerts will work reliably at scale. Plus, teens have found creative ways around age checks on other platforms, including Discord, and they can do the same with ChatGPT.
ChatGPT for Teens also risks normalizing age prediction and verification as a condition of accessing large language models (LLMs), where users of any age may be asked to surrender more information about their identity and behavior. That could put LLMs alongside social media platforms and adult websites as another category of online service driving the spread of age verification laws. As the regulation of AI develops, such as the EU AI Act, age verification is likely to become a standard requirement for chatbots, too.
A safer AI assistant that protects your privacy
Lumo is a different, private AI assistant. It doesn’t estimate your age based on how you use it, nor does it log your conversations or train on your data. We never scan or share your chats with anyone, a policy backed by the same zero-access encryption principles that protect the rest of Proton’s products.
Anyone 13 or older can use it, with parental consent required for minors, and there’s no profiling mechanism running beneath it. If you’re looking for a ChatGPT alternative built around privacy, you start there.
If you want a deeper look at how AI tools collect and use your data, our guide to AI breaks down the surveillance, security, and privacy practices.
