Governance and Regulatory Frameworks in Agentic AI

Governance and Regulatory Frameworks in Agentic AI
Proposer(s)
ANACOM, National Communications Authority of Portugal.
Summary
The emergence of agentic AI raises new governance and regulatory challenges as autonomous AI agents increasingly interact with humans, other agents, digital services and critical infrastructures, potentially exercising delegated authority and making decisions with legal, economic and societal consequences.
This theme aims to explore the governance principles and regulatory approaches needed to ensure a trustworthy, accountable, transparent and an effective human oversight in agentic AI ecosystems.

Scope / Open Questions
The exploration of this theme could address the following areas:
1. Governance principles and regulatory approaches
• What governance principles should underpin trust and identity frameworks for agentic AI?
• How can existing regulatory frameworks be adapted to address the specific characteristics of autonomous AI agents?
• How can risk-based regulatory approaches be applied?
2. Human oversight and fundamental rights
• How can meaningful human oversight and control be maintained as AI agents become increasingly autonomous?
• How can governance frameworks protect privacy, personal data, security and other fundamental rights?
• How should transparency and explainability requirements apply to interactions between humans and AI agents?

3. International governance and regulatory interoperability
   • What common principles could facilitate mutual recognition and cross-border trust?
   • How can regulatory fragmentation be avoided without undermining national regulatory autonomy?
   • What roles should governments, regulators, standardisation bodies, industry, academia and civil society play in shaping governance arrangements?

Related Work
This theme could build upon and complement existing international, regional and industry initiatives, including:
• United Nations: Global Digital Compact and relevant international initiatives on AI governance.
• European Union: AI Act, GDPR.
• OECD: OECD AI Principles and related work on trustworthy AI and AI governance.
• Council of Europe: Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law.

The objective would be to identify synergies, governance gaps and opportunities for international coordination, rather than duplicating existing regulatory or standardisation activities.

Discussion
This proposal is intended to initiate an open discussion on the governance and regulatory dimensions of trust and identity in agentic AI.
A central question is whether existing governance and regulatory frameworks are sufficiently equipped to address autonomous AI agents, or whether additional principles, mechanisms and institutional arrangements may be needed.

Possible outcomes could include a mapping of existing governance and regulatory approaches, identification of emerging challenges.
The scope and potential deliverables should remain open to discussion and refinement within FG-TIDA.