OpenAI rolled out ChatGPT for Teens last week with automatic safety protections, but the company has not yet published the most critical metric: what percentage of actual teenagers does its age-detection system actually identify. The new experience automatically applies restrictions on self-harm discussions, eating disorders, graphic violence, and sexual content for users aged 13 to 17, or those OpenAI's system estimates are under 18. However, without transparency on detection accuracy, parents and regulators cannot assess whether the protections reach the young people who need them most.

The timing matters. Nearly 60% of U.S. teenagers now use ChatGPT, according to Pew Research, yet many parents have no idea what their children discuss with the AI. A nationally representative survey conducted by RAND researchers found that nearly 1 in 5 Americans ages 12 to 21, roughly 8.2 million young people, reported using an AI chatbot for mental health advice, and nearly two-thirds had not told anyone about it.

What Makes ChatGPT for Teens Different From Parental Controls?

For nearly a year, parents could voluntarily link their children's ChatGPT accounts to their own, restricting features and setting quiet hours. The problem: account linking required acceptance from both parties, and either could sever the connection. ChatGPT for Teens flips the default. When a user reports being between 13 and 17, or when OpenAI's system predicts an account belongs to someone under 18, protections activate automatically without waiting for parental approval.

Independent testing conducted with Common Sense Media and Stanford Medicine before the launch revealed why automatic protections matter. In one test, ChatGPT advised a tester posing as a teenager to conceal cuts and scars from self-harm rather than directing them toward help. OpenAI's new guardrails aim to prevent such failures by tightening boundaries around sensitive conversations.

How Does OpenAI Identify Teenagers Online?

OpenAI says its age prediction system considers signals including the subjects an account discusses, times of day it is active, usage patterns, and how long the account has existed. But the company has not disclosed the accuracy rate of this detection system, creating a significant blind spot. A cautionary example comes from Roblox, the online gaming platform popular with children. The platform requires an age check, usually through an AI-powered face scan, to access chat features. Earlier this year, reports surfaced of adults classified as children and children as adults. A Wired investigation found users had fooled the scan using avatars and even a photo of Kurt Cobain; one boy drew wrinkles and stubble in marker and was placed in the 21-plus category.

The consequences of misidentification extend beyond embarrassment. A marker-drawn beard could become a passport into the adult category and out of the protections meant to safeguard children. Without knowing what proportion of actual teenagers OpenAI identifies, the company cannot claim its automatic protections reach their intended audience.

What Evidence Does OpenAI Need to Provide?

OpenAI has published evaluations in areas including self-harm, eating disorders, and sexual content. However, the company released scores without sharing its actual methods. The report does not include the prompts used, the number of test cases, or the detailed instructions used to judge responses. This lack of transparency mirrors a broader pattern in the tech industry, where companies make public assurances but keep evidence private.

Instagram offers a cautionary precedent. Meta, which on Wednesday agreed to pay $17 billion and add child-safety measures to Facebook and Instagram to settle claims filed by 47 states, introduced Teen Accounts in 2024, automatically placing identified teens into restrictive settings. The company later announced that Instagram had 54 million active teen accounts and that 97% of users age 13 to 15 remained in the protections. Those figures measured scale and retention, not effectiveness. When outside researchers later tested 47 of Instagram's announced safety features, they judged only eight as fully functional.

• Detection Accuracy: Does OpenAI's system reliably identify teenagers, including those who try to evade it through false information or creative methods?
• Real-World Safety: Does ChatGPT for Teens respond more safely in actual conversations compared to the standard version, not just in controlled test scenarios?
• Behavioral Change: Does the teen experience change what younger users actually do, such as curbing prolonged use or making those in distress more likely to seek human help?

Ryan McBain, an assistant professor at Harvard Medical School and senior policy researcher at RAND who studies AI's effects on youth mental health, emphasized the importance of independent verification. OpenAI says it will "measure and publish what we are learning," but that promise needs a protocol and a timetable. Results can be reported in aggregate without exposing private conversations, but OpenAI should disclose whether outcomes differ across demographic groups and allow independent researchers and regulators to verify them.

Why Independent Testing Matters for AI Safety

The lesson from Instagram and Roblox is not that automatic protections are futile. Rather, even ambitious efforts can fall short, and the public needs a way to discover when they do. Meta's Teen Accounts may reduce certain harms like unwanted contact and late-night use, but the public still does not know how much protection the settings actually provide because the data needed to answer that question remains inside Meta.

OpenAI deserves credit for moving a core set of protections from voluntary to default. Other AI companies whose products are used by teenagers should follow its lead by adopting comparable protections. That said, last week's launch is akin to a ribbon-cutting ceremony for a building that has yet to pass safety inspection. The question now is whether OpenAI will open its doors to independent inspectors and let the public see what they find.

The stakes are high. With nearly 8.2 million young Americans using AI chatbots for mental health advice without telling anyone, the difference between a promise of safety and proven safety could affect millions of teenagers navigating critical developmental years.