ChatGPT and Teenagers: Secure Access to AI in 2023

The Importance of Secure Access to AI for Teenagers

Today's teenagers are the first generation to grow up surrounded by artificial intelligence, a technology that will profoundly influence their future. Currently, about 90% of teenagers using ChatGPT do so for activities such as learning, researching information, developing skills, or enhancing their weekly productivity. Therefore, it is crucial that they have access to this technology. Restricting their access to AI until adulthood would be akin to depriving a previous generation of the Internet or search engines until they turned 18, leaving them less prepared to use a key technology of their time. However, this access must be framed by protective measures suitable for teenagers.

In a recent session of the OpenAI Academy, Oriana McKenzie illustrated how ChatGPT helped her excel in her high school studies. By optimizing her study methods, she was able to dedicate more time to her favorite sports and school activities. Her experience demonstrates the potential of AI to assist teenagers in learning and creating effectively.

To maximize these benefits, it is essential to implement robust protections tailored to their age. This includes automated safeguards that allow teenagers to explore, learn, and create safely while providing protections appropriate to their developmental stage. As AI progresses, it is our responsibility to ensure broad access while ensuring appropriate protections, enabling teenagers to leverage this evolving technology.

Over the past year, this has translated into strengthening default protections for teenagers, deploying age prediction tools, expanding parental controls, creating additional family resources to help parents support healthy and responsible use, and introducing learning features to promote deeper understanding rather than simply providing answers.

Our Key Commitments

Our approach is based on four fundamental commitments:

• Prioritize the safety of teenagers, even if it conflicts with other goals.
• Encourage real-world support when necessary.
• Treat teenagers as such, with realistic expectations.
• Be transparent by setting clear expectations.

Building for Learning, Not Just for Answers

Learning is one of the main benefits that teenagers can gain from AI. That’s why we are developing tools that encourage active engagement, critical thinking, and deeper understanding. The Study Mode, designed in collaboration with teachers, learning scientists, and pedagogy experts, helps students solve problems step by step through guiding questions, structured explanations, and opportunities for reflection, rather than simply providing the answer. Early assessments of this tool have shown promising improvements in student performance, influencing our research on how AI can support learning outcomes.

While teenagers can activate this mode themselves, parents with linked teenage accounts can now enable Study Mode directly through Parental Controls. When activated, it remains on by default whenever a teenager starts a new conversation, providing families with an additional way to guide the use of ChatGPT for homework and study.

We have also recently introduced education-focused starter prompts to help teenagers begin tasks such as breaking down a topic into simple steps, turning notes into a study guide, creating flashcards or practice questions, and checking for evidence and clarity.

We continue to expand interactive learning experiences. Research has consistently shown that people learn more effectively when they can actively engage with concepts rather than passively consume information. Since its launch earlier this year, 18 million weekly users are now interacting with interactive math and science experiences in ChatGPT, and we have expanded these experiences to over 250 new topics, ranging from integrals and mitosis to moon phases, photosynthesis, and more. We have also introduced a pronunciation experience that uses audio to help people learn to pronounce words in over 61 languages.

Protecting Teenagers with Built-in Safeguards

If our system estimates that a person using ChatGPT is under 18 years old, we automatically provide a more age-appropriate experience. Teenagers can still use ChatGPT to learn, create, and explore, but with additional protections designed to reduce their exposure to content that may not be suitable for them. This includes stricter safeguards regarding graphic violence, self-harm, risky viral challenges, unhealthy body image content, and dangerous, romantic, or sexual role-playing. The goal is to ensure that ChatGPT remains a tool for learning and creativity, not a substitute for real-life relationships.

We also design ChatGPT to encourage healthy habits. Teenagers who spend a lot of time on ChatGPT will now receive more frequent break reminders encouraging them to take a pause and step away. These reminders aim to help young people develop balanced technology habits while allowing them the freedom to use AI for learning, creativity, and problem-solving.

We believe that the best safety tools can help families navigate teenagers' online experiences together. We have built parental controls and notifications that allow them to set quiet hours, disable voice mode, manage access to image generation, and receive notifications in certain high-risk situations, such as indications of potential self-harm.

We are expanding these notifications to include cases where a linked teenage account has been disabled for violating our usage policies regarding violent threats or acts of online violence. This approach helps parents know when something serious has occurred while respecting teenagers' privacy and encouraging offline conversations and support.

As we developed these features, we consulted several experts, including Moonshot, a leader in online violence prevention, to ensure that these interventions are thoughtful, effective, and focused on helping families when needed.

Collaborating with External Experts

Protecting young people online is not something a single company can solve alone. We work closely with teenagers, parents, educators, child safety experts, mental health professionals, researchers, governments, and civil society organizations. These partnerships help us better understand the needs of young people, identify emerging risks, assess the effectiveness of our protections, and ensure that our policies are informed by independent expertise.

This work encompasses our approach to both learning and the well-being of young people. We collaborate with educators, including the American Federation of Teachers, students, and learning organizations to better understand how AI can support teaching and learning while enhancing active engagement and critical thinking. We also work closely with experts in mental health, child development, and youth safety, including the American Psychological Association, our Global Physician Network, and our Expert Council on Well-Being and AI.

Our work to build a healthier digital ecosystem goes beyond our own products. We have called for the development of scalable global standards for youth safety in AI through a dedicated AI Safety Institute, policies, and open-source tools that developers can integrate into their own AI systems, and to work with organizations such as Common Sense Media and everyone.ai to share learnings and best practices.

This year, we also joined the Family Online Safety Institute (FOSI), a long-standing advocate for safe online experiences for families and young people. We will work together to ensure that young people can benefit from AI, with empowered parents and responsible industry practices at the center. We hope this partnership will help advance the shared understanding of what works, strengthen appropriate age protections and parental tools, and support industry-wide efforts that enable young people to benefit from AI safely and responsibly.

Future Perspectives

Protecting young people online requires ongoing progress. In the coming months, we will continue to strengthen age-appropriate protections, give parents more tools and control, enhance protections against serious dangers, advance research on healthy AI use, and build more experiences that help teenagers learn actively and use AI with confidence.

We know there is still much to be done, and we are committed to continuing our work in collaboration with teenagers, parents, educators, experts, and communities around the world to provide broad and safe access for teenagers.