[Theme] Network-Native Governance and Trust Enforcement for Agentic AI

Theme Name

Network-Native Governance and Trust Enforcement for Agentic AI

Proposer(s)

Gianpaolo Angelo Scalone (Vodafone)

Summary

Current discussions on Agentic AI focus primarily on identity, authentication, authorization, attestation, and discovery. However, establishing trust is only the first step; trusted infrastructures must also be able to govern, constrain, revoke, monitor, and audit agent activities throughout their lifecycle.

This theme explores how network and telecommunications infrastructures can provide continuous trust enforcement and governance functions for autonomous agents, enabling accountable operation across enterprises, service providers, jurisdictions, and digital ecosystems. The objective is to identify architectural principles and trust models that extend beyond identity toward operational trust management.

Scope / Open Questions

What governance functions are required once an agent has been identified and authenticated?
How can delegated authority be constrained, monitored, and revoked throughout an agent's lifecycle?
How can accountability and traceability of agent actions be maintained across organizational and jurisdictional boundaries?
What mechanisms are needed for policy enforcement, risk management, and dynamic authorization of autonomous agents?
What role can telecommunications and network infrastructures play as trusted governance and enforcement layers for agentic ecosystems?
How should trust, delegation, authorization, observability, and enforcement interact within an overall Agentic AI trust architecture?
How can governance and enforcement operate consistently across heterogeneous trust domains?
What requirements exist for logging, auditing, dispute resolution, and forensic investigations involving autonomous agents?

Related Work

ITU-T XSTR.ATHENA (Autonomous Trusted Hierarchical ENtity Architecture) and related work on trust, identity, delegation, and governance for Agentic AI.
IETF OAuth 2.0, GNAP, and emerging agent authorization and delegation models.
OpenID Foundation and decentralized identity (DID/VC) ecosystems.
EU AI Act, GDPR, eIDAS 2.0, Data Act, and other governance-oriented regulatory frameworks.

Related Themes

This theme is intended to complement, rather than replace, several existing proposals:

Intent-based Security Policies (runtime interpretation and enforcement of delegated intent).
Remote Attestation for Agentic AI (governance after attestation has been established).
Cross-border Trust Management for Digital ID (policy enforcement across jurisdictions).
Agent Identity and Trust Stack (potential governance and enforcement layer within a broader trust architecture).
Sovereign Discovery (trust evaluation and policy enforcement following discovery).

Potential future consolidation with broader work on governance, delegation, and trust lifecycle management may be appropriate.