OpenAI calls for global action on youth AI safety through a dedicated AI Safety Institute
AI can be a powerful tool for young people. It can help students understand a difficult concept, practice a new language, prepare for a job interview, explore creative ideas, or learn a new skill. Just as the literacy movement spread across the globe in the middle of the 20th century required mass distribution of printed text and access to trained teachers, today providing access to safe AI will unlock new opportunities for personalized learning, help reduce barriers for the underserved, and increase workforce readiness.
Because the potential benefits are so significant, it is essential that young people can access AI in ways that are safe, age-appropriate, and supportive of healthy development. That responsibility should not fall primarily on parents or young people themselves. Companies, including OpenAI, have a responsibility to build products with appropriate safeguards by default, while empowering families with tools and information to guide how AI is used.
When Heads of State gather at the G7 Leaders’ Summit in Évian, France, later this month, the topic of youth AI safety will be a key focus. OpenAI will join to discuss greater collaboration on youth safety, how we can work together to raise industry standards, and OpenAI's new call for an international youth safety institute to be established.
The G7 Leaders’ Summit represents a rare opportunity for global coordination on youth safety. But a dedicated institute would provide the continuity and follow-through needed beyond a single summit, helping governments, researchers, civil society, and industry keep working together to share evidence, develop guidance, and raise standards over time.
There are several credible ways this could be done: through a new international institute, or by giving an existing or newly established national AI institute a global mandate to share research, evidence, and guidance with partners around the world. What matters most is the function it serves. Youth AI safety needs sustained attention, trusted evidence, and practical guidance that can keep pace with the technology.
Such an institute can build on emerging initiatives such as Common Sense Media’s Youth AI Safety Institute, supported by the OpenAI Foundation, as well as practical collaborations with educators, including OpenAI’s work with the American Federation of Teachers. It should also draw on real-world deployments, such as Estonia’s national ChatGPT rollout in schools, where OpenAI is working with Stanford and Estonian researchers to study impact and inform safer, more effective use in learning.
OpenAI welcomes the opportunity to work with the French Government, other G7 governments, and partners across civil society, academia, and industry on this vital work and to raise global standards on youth safety.
We believe strong AI youth safety frameworks, including any potential agreement reached at G7 convenings, should be led by principles that we are setting out here.
- Companies should know when a user is a minor and apply age-appropriate protections. This means requiring providers to use means such as effective, privacy-preserving age estimation to distinguish minors from adults, and defaulting to protective safeguards when a user’s age cannot be determined. Without this foundation, even the most well-intended youth protections may fail to reach the young people they are intended to protect.
- Companies should regularly assess both risks and benefits for young people and take steps to address them—before harm occurs. This means requiring providers to complete annual youth safety risk assessments and implement proportionate safeguards based on the risks they identify. Such assessments should consider risks to young people based on their developmental stage, empirical evidence from actual use, and relevant research or regulatory guidance. Assessments should consider not only potential harms, but also whether AI is supporting positive outcomes such as learning, creativity, skill development, and access to opportunity.
- Parents and guardians should have accessible, easy-to-use controls to help them guide their children’s experiences while supporting age-appropriate independence. These should include tools to manage key settings, including memory, data use, and time limits. Parental tools should be actively promoted so parents know they exist and understand how to use them.
- Families deserve clear, understandable information about how companies are protecting young users and provide meaningful information about both the benefits and risks of AI use by young people. Companies should publish safety policies that explain what safeguards are in place to protect young people, parental tools available, and how protections are updated as risks evolve. This transparency builds trust and helps parents make informed choices.
- Companies should have clear protocols in place to address serious safety situations including self-harm, exploitation, grooming, sexually exploitative content, and other high-risk interactions. These should include in-service support, referrals to appropriate resources, and timely parental notifications when appropriate. Companies should also design their systems to prevent them from generating content that is unsafe or developmentally inappropriate for kids, including graphic sexual or violent material.
- AI systems used by young people should be designed to support learning, development, and real-world relationships—not replace them. AI should serve as a tool that helps young people learn, create, build skills, and prepare for the future. At the same time, companies should establish clear boundaries in areas where healthy development depends on human judgment, real-world relationships, and professional support.
- The personal information of minors should be protected. This means prohibiting privacy-invasive, targeted advertising to young people, and barring companies from selling personal information.
- Youth safety frameworks should promote opportunity, literacy, and protection. Young people should have access to AI tools that support learning, creativity, skill development, and future opportunity, while also being equipped with the knowledge and critical-thinking skills needed to use AI safely and responsibly. This includes promoting AI literacy and helping young people understand the opportunities and limitations of AI systems, so they can engage confidently and responsibly in digital spaces.
- Strong accountability mechanisms, including independent audits, are essential to ensure these protections are meaningful in practice. Audits should be underpinned by a set of common standards that allow for interoperable audits across jurisdictions. Legislative frameworks should include oversight and enforcement measures that allow governments to assess whether companies are effectively implementing youth safety safeguards, mitigating identified risks, and complying with youth safety and privacy obligations.
The principles are reflected in how we build and operate ChatGPT, from model behavior and product design to expert input and real-world support. We have strengthened safeguards for users under 18, launched parental controls with proactive notifications, and advanced age-prediction systems so ChatGPT can apply stronger protections when someone may be under 18. Our Model Spec, which outlines how our models should behave, also includes dedicated principles for users under 18(opens in a new window), prioritizing teen safety, age-appropriate experiences, real-world support, and clear expectations. In practice, this means stronger guardrails around self-harm, dangerous activities, graphic content, body image, and secrecy, along with encouragement to seek trusted offline support or crisis resources when needed. When age is uncertain, we default to stronger safeguards.
AI designed with strong safeguards and offered with clear guidance, and support from families and educators, can help young people safely benefit from tools that expand learning, creativity, and opportunity. AI can help students understand difficult concepts, practice languages, improve their writing, learn to code, organize research, explore creative ideas, and prepare for jobs that are already changing. Through Education for Countries, OpenAI is working with governments and educators on research-driven deployments, localized learning tools, and teacher training so AI adoption is grounded in evidence and real classroom needs. This work is already supporting national education efforts with our partner countries including Estonia, Greece, and Singapore.
In addition to participation in the Leaders’ Summit, we will bring the OpenAI Forum(opens in a new window) to Paris to convene a practical conversation on how governments, researchers, civil society, educators, and industry can turn shared youth safety goals into concrete safeguards, standards, and implementation practices. France's Ambassador for AI and Digital Affairs, Clara Chappaz, will join OpenAI Chief Global Affairs Officer Chris Lehane, youth safety leaders from iRaise/Everyone.AI, and other experts for a discussion on how young people can benefit from AI while reducing risks.
Facts Only
OpenAI is calling for global action on youth AI safety through a dedicated AI Safety Institute.
The G7 Leaders’ Summit in Évian, France, will focus on youth AI safety, with OpenAI participating in discussions.
OpenAI proposes an international youth safety institute to provide continuity and collaboration beyond the summit.
The institute could be a new entity or an expansion of an existing national AI institute with a global mandate.
OpenAI supports initiatives like Common Sense Media’s Youth AI Safety Institute and partnerships with educators, including the American Federation of Teachers.
Estonia has rolled out ChatGPT in schools, with OpenAI and Stanford researchers studying its impact.
OpenAI outlines principles for youth AI safety, including age verification, risk assessments, parental controls, and transparency.
Companies should apply age-appropriate protections, assess risks, and provide tools for parents to manage AI use.
AI systems should support learning and development, not replace human judgment or real-world relationships.
OpenAI has implemented safeguards for users under 18, including stronger guardrails on self-harm, graphic content, and privacy.
OpenAI is working with governments and educators in Estonia, Greece, and Singapore on AI adoption in education.
The OpenAI Forum will convene in Paris to discuss youth safety goals with governments, researchers, and civil society.
Executive Summary
OpenAI is advocating for global action on youth AI safety, proposing the establishment of an international youth safety institute to ensure AI tools are safe, age-appropriate, and supportive of healthy development. The initiative will be discussed at the upcoming G7 Leaders’ Summit in Évian, France, where OpenAI will join discussions on raising industry standards and fostering collaboration among governments, researchers, and civil society. The company emphasizes that companies, not just parents or young people, bear responsibility for implementing safeguards by default while providing families with tools to guide AI use.
OpenAI outlines key principles for youth AI safety, including age verification, regular risk assessments, parental controls, transparency, and protocols for addressing high-risk interactions. The company highlights its own efforts, such as strengthened safeguards for users under 18, parental controls, and partnerships with educators and governments to integrate AI into learning environments. The proposal builds on existing initiatives like Estonia’s national ChatGPT rollout in schools and collaborations with organizations such as Common Sense Media. The goal is to balance opportunity and protection, ensuring AI supports learning, creativity, and workforce readiness while mitigating risks like exploitation or harmful content.
Full Take
OpenAI’s proposal for a global youth AI safety institute reflects a growing recognition of the need to balance innovation with protection, particularly for vulnerable populations. The strongest version of this narrative is that proactive, collaborative governance can mitigate risks while unlocking AI’s potential for education and opportunity. OpenAI deserves credit for advocating structural safeguards rather than relying solely on individual responsibility, and its principles—such as age verification, risk assessments, and parental controls—align with emerging best practices in digital safety.
However, the proposal raises questions about implementation and accountability. While OpenAI emphasizes industry responsibility, the effectiveness of self-regulation remains debated. The call for "strong accountability mechanisms, including independent audits" is critical but vague—who defines the standards, and how are they enforced? The reliance on partnerships with governments and civil society is promising, but without clear enforcement mechanisms, there’s a risk of symbolic action over substantive change. Additionally, the focus on "age-appropriate" protections assumes a universal standard, yet cultural and developmental contexts vary widely. The proposal’s success hinges on whether it can navigate these complexities without becoming a vehicle for industry self-interest or regulatory capture.
Root cause: This narrative echoes broader tensions in tech governance—balancing innovation with safety, global standards with local needs, and corporate responsibility with public oversight. The assumption that AI can be universally "safe" for youth overlooks the dynamic nature of technology and the diversity of young users’ experiences. The historical pattern of tech companies advocating for self-regulation while resisting binding oversight looms large here.
Implications: If implemented effectively, this could set a precedent for global AI governance, particularly in education. However, if it lacks teeth, it may serve as a PR shield for companies while leaving young users exposed to evolving risks. The focus on opportunity alongside protection is commendable, but without equitable access, AI could deepen existing divides.
Bridge questions: How can global standards accommodate cultural differences in definitions of safety and development? What mechanisms ensure that industry-led initiatives don’t prioritize corporate interests over youth well-being? How do we measure the long-term impact of AI on young users’ cognitive and social development?
Counterstrike scan: A coordinated influence campaign might frame this as a benevolent industry-led solution to preempt stricter regulation, using partnerships with educators and governments to lend credibility. The actual content, however, includes concrete proposals for accountability and transparency, which suggests a more genuine effort—though the lack of specific enforcement details leaves room for skepticism.
Patterns detected: none
Sentinel — Human
The text functions effectively as a policy brief, balancing high-level ethical goals with concrete operational requirements, suggesting careful human-directed synthesis of complex information.