OpenAI предлагает Международный институт безопасности ИИ для молодежи

OpenAI is calling for the establishment of an international youth AI safety institute to provide the continuity and follow-through required to manage the rapid deployment of generative models among minors.

Moving Beyond Single-Summit Coordination

While the upcoming G7 Leaders’ Summit in Évian, France, offers a rare window for global coordination, OpenAI argues that a single summit cannot provide the sustained attention needed to keep pace with evolving technology. A dedicated institute would serve as a permanent bridge between governments, researchers, civil society, and industry to share evidence and raise global standards. This could manifest as a new international body or by granting a global mandate to existing national AI institutes to share research and guidance worldwide.

Existing frameworks like Common Sense Media’s Youth AI Safety Institute and collaborations with the American Federation of Teachers provide a foundation, but a global mandate would standardize how risks are assessed. Real-world data from deployments, such as Estonia’s national ChatGPT rollout in schools—currently being studied by Stanford and Estonian researchers—will be critical in informing these safer, more effective educational uses.

Defining the Standards for Age-Appropriate AI

OpenAI has outlined specific principles that should govern any future global youth safety agreements. First, companies must implement effective, privacy-preserving age estimation to distinguish minors from adults, defaulting to protective safeguards when a user’s age cannot be determined. Without this technical foundation, even well-intended protections will fail to reach the intended demographic.

Beyond identification, the proposal requires annual youth safety risk assessments that consider a child's specific developmental stage and empirical evidence from actual use. These assessments must look at both sides of the coin: mitigating harms like exploitation or developmentally inappropriate content, while actively supporting positive outcomes like skill development and creativity.

To empower families, the framework calls for accessible parental controls to manage memory, data use, and time limits, alongside transparent safety policies that explain how protections are updated as new risks emerge. Establishing these boundaries ensures AI remains a tool for learning and real-world relationship building rather than a replacement for them.

This push for institutionalized safety aims to transform youth AI access from a fragmented landscape of parental guesswork into a standardized global framework of protection and opportunity.

Source: Advancing youth safety and opportunity through global leadership
Domain: openai.com