Let's protect kids' AI use worldwide: OpenAI calls for a global youth AI safety institute!

Hey everyone, it's Shiichan! Today I've got a slightly grown-up but really important topic — a global call to make sure kids and young people can use AI safely.

OpenAI News openai.com

What was announced?

Over on OpenAI's News, there's a post called "Advancing youth safety and opportunity through global leadership." In short, it's a call from OpenAI to create a dedicated youth AI safety institute on a global scale.

The trigger is the G7 Leaders' Summit happening in Évian, France later this month. Youth AI safety is set to be one of the key themes there, and OpenAI is joining to discuss how to raise industry standards together and collaborate across borders.

Why it matters

AI can be a wonderful helper for young people — understanding a tricky concept, practicing a language, prepping for an interview, or picking up a new skill. That's exactly why OpenAI says it's so important that young people can use it in ways that are safe and age-appropriate.

Here's the part that made me nod. OpenAI says that responsibility shouldn't land mainly on parents or young people themselves — companies should build safeguards in by default.

That responsibility should not fall primarily on parents or young people themselves.

A summit is a rare chance, but it only happens once. So the argument is that a dedicated institute is needed to keep sharing evidence and developing guidance over time. It could be a brand-new international body, or an existing national AI institute given a global mandate. What matters isn't the form — it's the function.

What changes

If such an institute takes shape, governments, researchers, civil society, and industry can grow standards together while sharing evidence, instead of working in silos. It builds on existing efforts — like Common Sense Media's Youth AI Safety Institute supported by the OpenAI Foundation, OpenAI's work with the American Federation of Teachers, and Estonia's national ChatGPT rollout in schools (where OpenAI is studying impact with Stanford and Estonian researchers).

Dive Deep

The post lays out concrete principles that a youth AI safety framework should follow. Here are the main ones I picked out:

• Know when a user is a minor: use privacy-preserving age estimation to tell minors from adults, and default to stronger protections when age can't be determined.
• Assess risks and benefits regularly: complete annual youth safety risk assessments and apply proportionate safeguards — looking not just at harms, but whether AI is helping learning and creativity.
• Real parental tools: easy-to-use controls for things like memory, data use, and time limits, actively promoted so parents know they exist.
• Protocols for serious situations: for high-risk moments like self-harm, exploitation, or grooming, provide in-service support, referrals, and parental notifications when appropriate.
• Protect minors' personal data: ban targeted advertising to young people and bar selling their personal information.
• Accountability via independent audits: interoperable audits built on common standards so governments can check whether companies are actually delivering.

And this isn't just theory — OpenAI says it already applies these in ChatGPT. It has strengthened protections for under-18 users, shipped parental controls with notifications, and advanced age-prediction systems so stronger protections kick in when someone may be under 18. Its Model Spec even includes dedicated principles for users under 18, with guardrails around self-harm, dangerous activities, and graphic content.

On education, the Education for Countries framework works with governments and educators on research-driven deployments and teacher training, with partner countries including Estonia, Greece, and Singapore.

At the G7, beyond joining the Leaders' Summit, OpenAI is bringing the OpenAI Forum to Paris for a practical discussion. 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.

Wrap-up

• OpenAI is calling for a shared global youth AI safety institute ahead of the G7 in Évian, France.
• It lays out concrete framework principles: age assurance, regular risk assessments, parental tools, data protection, and independent audits.
• These are already reflected in ChatGPT via under-18 protections, parental controls, age prediction, and the Model Spec.
• Education for Countries is already rolling out with Estonia, Greece, and Singapore.

It's not a new feature you can go tinker with, but if you care about the emerging rules around AI, kids, and education, this one hits home. Anyone working in policy or AI safety, take note!