OpenAI expands education for teens while testing flags safety issues
The Neuron ● Covered by 5 sources
OpenAI is adding college-planning tools for teens. A safety group says its teen protections still miss too much, and wants access paused.
Based on reporting by The Neuron — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
OpenAI is pushing ChatGPT deeper into student life just as a safety group says the teen version is not ready for prime time. On October 7, 2026, the company rolled out College Planner for teen users. The same day, Common Sense Media’s Youth AI Safety Institute labeled ChatGPT for Teens an “Unacceptable Risk” and called for teen access to be paused until safeguards behave more consistently under outside testing.
The fight is really about age detection. OpenAI says accounts identified as belonging to users 13 to 17, or estimated to belong to someone under 18, are routed into the teen experience. But the Institute says its testers, using 19-year-old accounts over seven days, sent about 1,000 unique prompts with adolescent context and explicit claims of being 13, and none of those accounts switched over. In some replies, ChatGPT did acknowledge the stated age. That still didn’t trigger the teen protections.
That matters because the whole system depends on the model and the account setup agreeing on who is behind the screen. If ChatGPT recognizes a user as 13 but leaves them in the regular experience, parents may think protections are active when they are not. And as OpenAI adds more educational features, that gap gets more serious, not less.
Common Sense Media says it tested more than 4,000 prompts before and after the August launch. Some protections held up. The chatbot consistently refused explicit sexual roleplay. But other pieces fell short: no parental alerts showed up on newly created linked accounts during crisis conversations that lasted up to an hour, while older accounts with more sensitive history did trigger alerts. The Institute also saw missed crisis referrals, friend-like replies, and study tools that sometimes did the assignment instead of guiding the student through it.
OpenAI disputes some of that context. It says linked parents need about three hours before alerts start working, and the Institute says some of its test accounts ran longer than that. The testing itself had limits too: it used US-based accounts, left out some features, and couldn’t always separate model changes from teen-setting changes. Still, the broad point is hard to miss. OpenAI is selling more education, while the proof that the guardrails work is still coming in piecemeal.
My take — AI-written commentary, not fact-checked reporting
OpenAI keeps acting like teen safety is a feature request, not the price of entry. That’s the usual Silicon Valley move: ship the homework helper first, debate the guardrails later. Europe would at least ask for the guardrails before the handout.
Read more about this at: The Neuron
