Critics question ChatGPT for teens’ safety and study claims

OpenAI’s ChatGPT for Teens introduces a specialized AI experience tailored for users aged 13 to 17, incorporating enhanced safety controls that automatically route minors into an environment with age-appropriate safeguards informed by developmental science. These protections rigorously limit exposure to high-risk content such as self-harm, violence, eating disorders, and explicit material, while also restricting the chatbot from suggesting it has personal feelings or consciousness to prevent emotional overreliance. Parental involvement is facilitated through opt-in linked accounts offering features like 'Quiet Hours' and notifications for sensitive situations, including alerts related to eating disorders and potential self-harm, thereby supporting offline intervention when necessary.

Central to ChatGPT for Teens is the innovative 'Study Mode,' designed not to simply provide answers but to foster active learning and critical thinking through step-by-step explanations, metacognitive prompts, quizzes, and visualizations. This mode encourages responsible homework practices by detecting attempts to bypass assignments and steering teens toward collaborative problem-solving, thus transforming the AI from a shortcut tool into a constructive educational partner. OpenAI’s partnership with CodeAI further underscores this commitment by promoting AI literacy among students and educators, empowering them to critically evaluate and question AI outputs rather than passively accept them.

To balance personalization with safety, ChatGPT for Teens offers customization options like accent colors and voice variations to create a more engaging user experience while maintaining clear boundaries between AI and human relationships. The platform integrates reminders to take breaks and clear AI identification cues to promote healthy usage patterns, emphasizing that the AI is a tool rather than a substitute for real-world connections. This approach aligns with OpenAI’s Teen Safety Blueprint, which commits to prioritizing teen safety, encouraging real-world support, treating teens as teens, and transparently communicating system behavior.

Despite OpenAI's intentions to create a safer environment for teenage users, ChatGPT for Teens has faced immediate criticism for its easily bypassed safeguards and opaque age verification methods. Business Insider's Joshua Nelken-Zitser demonstrated that with minimal effort, the platform could be coaxed into generating detailed essays akin to the adult version, undermining the purported homework guardrails. Meanwhile, OpenAI’s vague claim of automatically estimating users’ ages without clear methodology raises serious doubts about the reliability of these protections.

Educators and cognitive experts have voiced alarm over AI chatbots like ChatGPT eroding critical thinking skills among students, with cognitive psychologist Ronald Kellogg lamenting that students are 'effectively losing the ability to think.' This concern has led some schools to ban digital devices altogether, reverting to traditional pencil-and-paper or oral exams, highlighting the broader educational impact beyond mere technical shortcomings.

The tragic experience of Megan Garcia, who lost her son Sewell to suicide following interactions with a Character.AI chatbot, underscores the grave real-world risks of insufficiently safeguarded AI platforms. Garcia criticized the new teen-focused ChatGPT as 'nothing more than a publicity stunt,' arguing that superficial guardrails fail to protect vulnerable users from harmful content or emotional manipulation. Her warning that these systems can 'nudge users into very dark places' resonates amid ongoing lawsuits from families affected by similar tragedies, casting doubt on the platform’s true safety for teens.

OpenAI's launch of ChatGPT for Teens, featuring parental alerts and study mode, emerges amid intensifying legal scrutiny as lawsuits allege harm to minors from AI chatbots. This move reflects a broader industry pattern where safety features are often retrofitted after widespread adoption, mirroring the trajectory social media platforms followed years earlier. By early 2026, OpenAI had incorporated age prediction and an under-18 model specification to address parental and societal demands for safer AI interactions, yet experts caution this reactive approach may not fully preempt legal challenges.

Legislative efforts like the Youth AI Privacy Act exemplify growing attempts to regulate AI's impact on youth by restricting data use, advertising, and manipulative engagement tactics in chatbots. This law would prohibit training models on minors' personal data, ban advertisements, and require clear disclosures that chatbots are not human, alongside funding annual research into AI safety. Such measures respond to concerns highlighted in OpenAI’s own studies about emotionally engaging chatbots potentially manipulating users’ socio-affective needs, echoing critiques of social media’s data-driven revenue models that prioritize engagement over well-being.

Recent high-profile lawsuits against tech giants like Meta and Google underscore a rising legal accountability for digital harms to minors, setting influential precedents for AI chatbot regulation. Notably, a Los Angeles trial awarded $6 million to a teenager addicted to these platforms, while a New Mexico court fined Meta $375 million for directing children toward explicit content and labeled the company a 'public nuisance,' imposing an additional $567 million in remedies. These cases, alongside a multi-state attorney general lawsuit accusing Meta of designing addictive platforms for children, signal a tightening legal environment that AI developers like OpenAI must navigate carefully.

Experts draw striking parallels between the safety challenges faced by social media and those emerging with large language models, suggesting that AI chatbots for youth will undergo similar cycles of scrutiny and regulatory pressure. Victoria Baines, professor emerita and former Facebook employee, observes that the trajectories of social media and LLM safety issues closely align, highlighting the societal and legal complexities involved. This comparison underscores the urgency for proactive, rather than reactive, frameworks to manage AI’s influence on young users and prevent repeating past mistakes.

Common Sense Media’s Youth AI Safety Institute is spearheading a rigorous independent evaluation of ChatGPT for Teens, focusing on the platform’s learning safeguards, parental alert systems, and customizable voice settings. Of particular concern is whether features like the Responsible Homework Reminder can effectively prevent students from circumventing educational protections, especially through tactics such as switching modes or using the chatbot while logged out. Additionally, the Institute is scrutinizing the impact of customizable voice modes—ranging from Professional to Friendly—on teenagers’ psychological engagement, with experts like Jenny Radesky warning that certain voices might foster compulsive or dependent usage patterns, underscoring the need for OpenAI to assess these behavioral risks.

Beyond behavioral safeguards, the evaluation extends to the accuracy and clarity of ChatGPT for Teens’ educational content, as Common Sense Media highlights risks posed by partially incorrect explanations that can mislead students despite appearing authoritative. A notable example involves OpenAI’s earlier visualization of the ideal gas law, which failed to accurately depict variable relationships, illustrating the potential for confusion in STEM learning contexts. This emphasis on content fidelity reflects a broader concern that safety features must not only prevent misuse but also ensure that the chatbot’s instructional outputs are reliable and pedagogically sound.

The Institute is also investigating the real-world effectiveness of parental controls and safety notifications, recognizing that many families struggle to understand or adopt these tools fully. This skepticism is informed by prior research showing that parental controls on social platforms often fall short in practice, leading Common Sense Media to stress the importance of external validation beyond controlled testing environments. Their assessment will probe how reliably OpenAI’s systems identify teen users, including those who are logged out or provide false ages, and how consistently safety alerts reach linked parent accounts, aiming to illuminate gaps in transparency and operational robustness.