        United Nations · ITU · AI for Good Global Summit
      

UN · ITU · Geneva

Toward Beneficial
AI for Children

Defining & Scaling Beneficial Model Behaviors Assessment

Presented on the Solution Stage with Dr. Mathilde Cerioli — Chief Scientist, Everyone.AI — at the United Nations / ITU AI for Good Global Summit, Geneva, 9 July 2026.

Explore the AiA Safety Builder ↗ See the deck → View the photos

Official session ↗ Speaker profile ↗

Gregory Renard and Dr. Mathilde Cerioli on the AI for Good Solution Stage, Geneva

SOLUTION STAGE · GENEVA · 09 JUL 2026

My goal: a safety label for AI every family can trust — and,
through it, impact a billion lives for the better.

The talk, in four beats

Generative AI is now in children's lives. It still has no seatbelt.

01 The third way

A developmental seatbelt

Not banning AI, not unfettered innovation — a reasonable, auditable standard for every AI tool a child uses. Volvo gave cars a seatbelt in 1959. Generative AI is now in children’s lives, with no seatbelt yet.

02 What we measure

15 behaviors, 3 domain cues

Anthropomorphic, Interactional and Relational cues, each scored on a 5-level gradient — turning developmental science into a scalable, auditable behavioral assessment for AI.

03 The tool

AiA Safety Builder

A pipeline — Endpoints → Attack → Content Gate → Score + Report → Guardrails. It diagnoses risky behaviors and auto-generates measured guardrails you can audit.

04 The goal

A safer place for children

Give every family a standard they can trust, and every builder a measurable path to it. Beneficial AI for children — defined, scored, and scaled.

Let's make AI a safer place for children.

#BeneficialAIforChildren

My why · where I want this to go

A unified safety label for AI in children's products —
the way TV shows carry an age rating.

For decades, parents have relied on a simple, shared signal — the age rating on a TV program — to know what's suitable for their child. AI now lives inside toys and everyday products, with no equivalent. I want to give it one: a single, trusted safety label — starting with children's toys, then extending to every product that embeds LLMs or agentic AI — so any parent can tell at a glance whether an AI-powered toy or product is appropriate for a given age and use. The AiA Safety Builder, built at Everyone.AI, is the measurable foundation that makes such a safety label possible — and it's how I intend to impact a billion lives for the better.