OpenAI Youth AI Safety Proposal and Global Leadership Initiative

OpenAI Youth AI Safety Proposal and Global Leadership Initiative

OpenAI is calling for the creation of an international youth safety institute and the adoption of a global framework for protecting minors using AI. This initiative aims to standardize age-appropriate safeguards and ensure that AI serves as a tool for learning and opportunity rather than a replacement for human development.

Proposal for an International Youth Safety Institute

OpenAI proposes the establishment of a dedicated international youth safety institute to provide continuity in global coordination beyond individual summits. This institute would serve as a hub for governments, researchers, civil society, and industry to share evidence, develop guidance, and raise safety standards over time.

OpenAI suggests two primary implementation paths for this institute:

  • A new international organization: A standalone entity dedicated to youth AI safety.
  • An expanded mandate: Granting an existing or newly established national AI institute a global mandate to share research and guidance with international partners.

This proposal builds upon existing efforts, including Common Sense Media’s Youth AI Safety Institute (supported by the OpenAI Foundation), collaborations with the American Federation of Teachers, and research-driven deployments such as the national ChatGPT rollout in Estonian schools conducted in partnership with Stanford and Estonian researchers.

Nine-Point Framework for Global Youth AI Safety Standards

OpenAI has outlined nine core principles to guide the development of strong AI youth safety frameworks and G7 agreements:

  1. Age-Appropriate Protections: Providers must use privacy-preserving age estimation to identify minors and default to protective safeguards when a user's age is undetermined.
  2. Proactive Risk Assessment: Companies should conduct annual youth safety risk assessments to address potential harms before they occur, while also measuring positive outcomes like learning and creativity.
  3. Parental Controls: Guardians should have accessible tools to manage memory, data use, and time limits, with these tools being actively promoted to families.
  4. Transparency: Companies must publish clear safety policies explaining the safeguards in place for young users and how those protections evolve.
  5. High-Risk Protocols: Clear protocols must exist for serious safety situations, including self-harm, exploitation, grooming, and sexually exploitative content, including referrals to resources and parental notifications.
  6. Support for Development: AI should be designed to support learning and real-world relationships, not replace them, maintaining boundaries where human judgment and professional support are essential.
  7. Privacy Protections: The personal information of minors must be protected, specifically prohibiting the sale of personal data and privacy-invasive targeted advertising.
  8. Promotion of Literacy: Frameworks should promote AI literacy and critical-thinking skills so young people can use AI responsibly and confidently.
  9. Accountability and Audits: Independent audits based on common, interoperable standards and legislative oversight are necessary to ensure protections are implemented effectively.

Implementation in ChatGPT and Global Education

OpenAI integrates these principles into ChatGPT through specific product designs and behavioral guidelines:

  • Safeguards for Minors: Strengthened guardrails around body image, dangerous activities, graphic content, self-harm, and secrecy for users under 18.
  • Model Spec: Dedicated principles for users under 18 that prioritize age-appropriate experiences and encourage seeking offline support.
  • Age Prediction: Advanced age-prediction systems that apply stronger protections when a user is suspected to be a minor.

Beyond product safety, OpenAI's "Education for Countries" initiative works with governments in Estonia, Greece, and Singapore to provide localized learning tools, teacher training, and research-driven deployments to ensure AI adoption meets actual classroom needs.

G7 Summit and the OpenAI Forum

OpenAI will participate in the G7 Leaders’ Summit in Évian, France, to discuss industry standards and the proposed youth safety institute. Additionally, OpenAI will host the OpenAI Forum in Paris, featuring discussions with France's Ambassador for AI and Digital Affairs, Clara Chappaz, and OpenAI Chief Global Affairs Officer Chris Lehane, alongside experts from iRaise/Everyone.AI.

Sources