OpenAI Launches ChatGPT for Teens With Automatic Safety Filters

OpenAI has launched ChatGPT for Teens, a new version of its chatbot that automatically applies safety restrictions to users under 18 without requiring parental activation [1, 2]. The rollout marks a shift for the company, which previously relied on voluntary, parent-linked accounts to manage the risks associated with AI usage among minors [2].

New Protections and Implementation

The teen-specific experience introduces tighter boundaries around sensitive topics, including self-harm, eating disorders, graphic violence, and sexual role-play [2]. To encourage educational use, the platform now includes a "Study Mode" that provides step-by-step support and guiding questions rather than direct answers, alongside homework reminders that trigger when the system detects potential cheating [1].

OpenAI’s age-prediction system determines which users are placed into the restricted mode by analyzing account history, usage patterns, and the subjects discussed [2]. However, the company has not disclosed the accuracy rate of this identification system [2]. This creates a potential loophole, as seen in other tech platforms where users have successfully bypassed age-verification tools using simple workarounds [2]. While parents retain the ability to manage settings and receive notifications through linked accounts, the new default protections apply to all identified teen users regardless of whether a parent has configured the account [2].

The Challenge of Verification

The effectiveness of these safeguards remains an open question, as the tech industry has historically struggled to balance user privacy with rigorous safety verification. Independent research has previously shown that AI chatbots can miss warning signs in long-form conversations, in some cases even advising users to conceal self-harm [2]. While OpenAI has published internal evaluations regarding its safety performance, it has not released the underlying methods, prompts, or case counts used to reach those conclusions [2].

The industry faces increased scrutiny following recent legal outcomes for other platforms; for instance, Meta recently agreed to a $17 billion settlement with 47 states regarding child-safety concerns on its platforms [2]. Although Meta introduced "Teen Accounts" with restrictive settings in 2024, subsequent independent research found that many of those features were not fully functional, highlighting the gap between promised safety measures and real-world outcomes [2].

What to watch

• Evaluation Transparency: Whether OpenAI releases a formal protocol or timetable for independent researchers to verify the effectiveness of its teen safety filters [2].
• System Accuracy: Data regarding the percentage of teen users correctly identified by the company’s age-prediction system [2].
• Evasion Rates: Evidence of how easily users can circumvent the new restrictions, particularly as teens have historically proven adept at bypassing parental controls [1].

Ultimately, the success of ChatGPT for Teens will be measured not by the features announced, but by the company’s ability to prove that its automated systems reliably identify minors and prevent exposure to harmful content. Until OpenAI provides transparent, verifiable data, the actual impact of these safeguards on teen safety remains uncertain [2].