Parent / guardian Child profile AI agent Child-Aware Delegated Identity for Agentic AI Privacy-preserving, age-aware AI agents Gianpaolo Angelo Scalone  |  Vodafone Core question: How can an AI agent know that it represents a child, and safely preserve that protection across services? 1
WHY NOW The protection gap moves with the child, but identity does not A child can move across devices, AI agents and services while each interaction starts again as an unknown or generic user. AI AI tutor AI Game agent AI Shopping agent AI Companion Age repeated Consent fragmented Rules inconsistent No portable revocation The missing capability Recognise a child profile Carry parental authority Minimise data disclosure Apply age-appropriate policy Revoke or update centrally 2
PROPOSED TRUST MODEL The concept: delegate protection, not personal data A trusted root issues a scoped, privacy-preserving child-profile credential that an agent can verify and enforce. 1  Trusted root Telco identity + accountable guardian 2  Delegation Child profile + parental authority + scope 3  Minimal proof Age band / protection class, not full identity 4  Agent policy Age-appropriate rules and permissions 5  Control Audit, update, revoke Lifecycle feedback: update • suspend • revoke • evidence The agent receives only what it needs to behave correctly. 3
ILLUSTRATIVE USE CASE One profile, consistent protection across an AI journey 1 Parent sets a child profile Age band, permissions, spending and content limits 2 Telco identity delegates Signed, scoped and privacy-preserving credential 3 AI tutor verifies Applies age-appropriate interaction and data rules 4 Shopping agent inherits limits No independent re-collection of full identity 5 Guardian changes policy Update or revocation propagates Outcome: child protections become portable policy, while identity disclosure remains minimal and controlled. 4
DESIGN PRINCIPLES Five principles to keep the theme rights-respecting 1 Child first Best interests and age-appropriate safeguards 2 Data minimisation Prove the needed attribute, not the person 3 Delegated authority Make guardian scope explicit and auditable 4 Portability Carry protections across agents and domains 5 Agency and remedy Support correction, appeal and human escalation Not a universal child identifier The theme should explicitly avoid persistent tracking and unnecessary identity correlation. 5
WHAT TIDA CAN DEFINE Proposed scope for a TIDA study theme Core architecture questions Governance and safeguards Identity and attribute model for a child profile Delegation from guardian or accountable authority Age-band and protection-class assertions Consent, scope, expiry and revocation semantics Cross-agent and cross-domain portability Privacy and unlinkability requirements Auditability, evidence and accountability Conflict handling across laws and policies Human escalation and remedy conditions Interfaces with AI policy and enforcement layers 6
The decision requested from TIDA Open a dedicated theme on child-aware delegated identity for Agentic AI. 1 Recognise the gap Age assurance alone does not make agent delegation child-aware. 2 Frame the study Define requirements, safeguards and interfaces, not a product. 3 Connect the work Link agent identity, trust enforcement, human oversight and child protection. Proposed next step: adopt and refine “How can an AI agent reliably protect a child without needing to know who the child is?” 7