[Theme] Global Trust and Accountability for Agentic AI

Theme Name

Global Trust and Accountability for Agentic AI

Proposer(s)

Bo Fjelkner (Ericsson)

Summary

Agentic AI is rapidly evolving from a tool used by humans into an active participant in digital ecosystems, capable of communicating, acting, and making decisions on behalf of individuals, organizations, and public authorities. While many jurisdictions are beginning to establish frameworks for digital identity, there is currently no globally aligned understanding of how trust, identity, accountability, delegation, and responsibility, and liability should be managed when autonomous agents interact with humans and other agents.

This theme would explore the need for global governance principles and policy guidance that can help governments, regulators, standards bodies, and industry develop interoperable and trustworthy approaches for the emerging agentic ecosystem.

Scope / Open Questions

Potential areas for exploration include:

• What constitutes a trusted identity for an AI agent?
• How should authority be delegated from a human, organization, or public authority to an AI agent?
• What information is required to establish trust between unknown agents operating across jurisdictions?
• How can accountability and liability be understood when AI agents perform actions on behalf of others?
• What mechanisms are needed to support transparency, auditability, and dispute resolution?
• How should trust and identity frameworks address cross-border interactions involving multiple legal and regulatory environments?
• Which policy and governance principles should be considered by jurisdictions that seek to regulate agentic AI ecosystems?
• What role could trusted digital identity infrastructures play in supporting trustworthy agent interactions?
• How can global interoperability be achieved while respecting regional legal and cultural differences?

Related Work

• ITU-T trust and identity initiatives
• UN Global Digital Compact
• ISO/IEC digital identity and AI standards
• W3C Decentralized Identifiers (DID) and Verifiable Credentials (VC)
• OpenID Foundation trust and identity specifications
• IETF work related to identity, trust, authentication, and attestation
• EU Digital Identity Wallet (eIDAS 2.0)
• EU AI Act
• OECD AI Principles
• National and regional AI governance initiatives

Related Themes

Potential overlaps may exist with:

• Digital identity for AI agents
• Delegation and representation models
• Trust frameworks and trust infrastructures
• Agent-to-agent communication and interoperability
• Governance and accountability models for agentic ecosystems

Discussion

The emergence of agentic AI raises fundamental questions that extend beyond technology and standards into governance, policy, and law. Unlike traditional digital systems, AI agents may increasingly act autonomously while representing the interests of humans, enterprises, and public institutions. This creates new challenges around trust, accountability, liability, and jurisdiction that are unlikely to be solved by any single country or region acting alone.

FG-TIDA could provide a neutral global forum to identify common principles, terminology, and policy considerations that assist governments, regulators, industry, and standards organizations in developing trustworthy and interoperable approaches for the emerging agentic ecosystem.