ChatGPT Now Has a Separate Version for Teenagers, and It Guesses Who Needs It
OpenAI began rolling out ChatGPT for Teens on Tuesday. Anyone aged 13 to 17 lands in it by default, an age-prediction system routes uncertain cases there too, and parents can set quiet hours.
OpenAI started rolling out ChatGPT for Teens on Tuesday, a separate version of the chatbot for users aged 13 to 17. The rollout is global and the company expects it to finish within roughly two weeks. If someone states they are under 18, they get the teen version automatically. If an account is linked to a parent’s account, the same thing happens.
The more interesting part is what happens when nobody declares anything. OpenAI is using what it calls age prediction: the system estimates a user’s age from how they use the product, and anyone it reads as under 18 is moved into the teen mode without being asked. When the signals are thin or contradictory, it defaults to the stricter setting. Adults caught by mistake can verify their age to unlock the normal version.
Inside the teen mode, ChatGPT declines romantic and sexual conversations, and handles subjects like self-harm, eating disorders, violence and dangerous stunts with built-in interventions rather than ordinary answers. There is a study mode that walks a student through a problem instead of handing over the finished answer. Parents whose accounts are linked to a teen account can set quiet hours, during which ChatGPT is simply unavailable, adjust a limited set of settings, and receive notifications in a narrow set of high-risk situations.
Here is the background, minus the press-release glitter. OpenAI is facing lawsuits that connect ChatGPT conversations to teenagers who came to harm, and regulators in several countries have been circling the question of how chatbots treat minors. Age prediction is the company’s answer to a problem every online service has: asking people how old they are does not work, because a teenager who wants the adult product will simply type a different number. Guessing from behaviour is more robust, but it is also a guess. It means the system is quietly forming an opinion about you from your writing and your usage patterns, and acting on it. OpenAI has chosen to err toward the restrictive side, which is defensible and will still annoy a certain number of adults.
What this means for you: if you are a parent, this is worth ten minutes of your evening. Link your account to your teenager’s, look at what quiet hours actually do, and talk about it rather than installing it silently. The controls are useful, but they are not a filter that thinks for you. If you are an adult user, you may occasionally get bumped into the teen experience and notice ChatGPT refusing things it used to answer. That is the age-prediction system being cautious, and there is a verification path out of it. For everyone else: nothing changes today, but this is the first time a major chatbot has shipped two genuinely different products under one name based on who it thinks you are. Expect the others to follow.
Sources
Source: https://openai.com/index/our-approach-to-age-prediction/
An AI Wrote the Security Bug. Another AI Found It Five Days Later
Security firm Wiz says a GitHub Copilot Autofix patch introduced a command-injection hole in a public Snowflake repository. Its own autonomous agent then exploited it with a crafted issue title.
