1. Rationale
Digital ecosystems are undergoing a structural transformation driven by the emergence of agentic Artificial Intelligence (AI) systems. An Agentic AI or AI Agent or Agent is an autonomous software entity that interprets goals, formulates intent, and executes actions, often by invoking external tools, APIs, or other agents to achieve outcomes on behalf of users. Its behavior is driven by model-based reasoning, typically relying on Large Language Models (LLMs) or other Foundation Models (FMs) to analyze context, generate plans, and adapt its actions dynamically.
These systems are no longer limited to passive computation; they now act as autonomous entities capable of reasoning, planning, and executing tasks across services, infrastructures, and organizational boundaries. As a result, they are increasingly integrated into critical digital workflows and interact directly with existing identity and trust infrastructures.
This evolution fundamentally changes the nature of identity in digital systems. Identity is no longer limited to human users and static machine instances but must also encompass autonomous, dynamic, and goal-oriented agents. Industry and research communities increasingly recognize that such agents must be treated as first-class entities within identity and access management systems, requiring dedicated identity models, lifecycle management, and governance mechanisms.
At the same time, separating human and agentic identity into distinct domains is neither practical nor desirable. Agentic AI systems act on behalf of humans, interact with human-centric services, and must remain ultimately accountable to human stakeholders. Identity relationships therefore span human-to-human, human-to-agent, and agent-to-agent interactions, requiring a unified and coherent approach to identity and trust management.
However, identity alone is insufficient. In agentic environments, the core question extends beyond "what an entity is" to "whether, and under what conditions, that entity should be trusted to act". Trust becomes a dynamic property that must be continuously evaluated, taking into account behaviour, context, delegation, and technical-policy compliance. This is particularly critical where agentic systems operate across organizational and jurisdictional boundaries, in which trust cannot be established or enforced by a single authority.
Existing digital identity and trust frameworks, including those developed within ITU-T and other standards development organizations (SDOs), have been designed primarily for human users or static services. While they provide a strong foundation, they do not fully address the requirements introduced by autonomous agents, such as dynamic delegation, continuous trust evaluation, and multi-agent interaction at scale. This results in fragmentation, inconsistent trust decisions, and limited interoperability across ecosystems.
These limitations are increasingly affecting the deployment of agentic AI systems. Despite significant investment and rapid innovation, organizations face challenges in scaling such systems owing to gaps in trust, governance, and interoperability. The absence of common frameworks for identity, trust evaluation, and accountability remains a key barrier to the transition from experimental deployments to reliable, large-scale adoption.
In this context, there is a clear need for a coordinated pre-standardization effort to establish common terminology, reference architectures, trust models, and interoperability mechanisms for digital identity and trust management in environments involving both humans and agentic AI.
The establishment of this Focus Group provides a platform to bring together stakeholders from industry, academia, and standardization communities to build a shared understanding of these challenges, identify gaps in existing approaches, and define the foundations for future standardization. This work will support the development of interoperable, secure, and trustworthy digital ecosystems in which humans and agentic AI can safely interact and collaborate.
In particular, the Focus Group will foster collaboration with external stakeholders and relevant standardization fora to ensure coherence and avoid duplication of effort. When potential deliverables are identified, early consideration shall be given to the most appropriate organizations or fora for their further development, facilitating an effective transition from pre-standardization activities to formal standardization work.
2. Scope
The Focus Group studies trust management and interoperable digital identity infrastructure for humans and for agentic AI including when leveraged by embodied AI. The Focus Group will explore transforming into a Collaboration platform on Trust Management (e.g. similar to CITS).
The Focus Group examines existing and emerging technologies, standards, and implementation approaches, with the aim of identifying gaps and supporting pre-standardization activities. It considers both human-centric and agentic AI systems, including their interaction, and explores how identity management and trust can be operationalized in multi-actor ecosystems involving users, agents, service providers, and relying parties.
The Focus Group will identify stakeholders with whom ITU-T could collaborate and will enable the inclusion of non-members to contribute to the technical pre-standardization work. Potential collaboration may be developed across academia, industry, regulatory bodies, and standards development organizations (SDOs), to ensure inclusive and interoperable technical foundations.
The following topics are out of scope for the Focus Group:
- Agentic AI protocols
- AI governance
- Digital ID (and its content) as this is a National matter and outside the scope of ITU
- Considering the dynamic nature of this emerging field, particular attention will be paid to the A.7 clause « The subject is not already addressed by work underway in ITU-T study groups or other focus groups, or cannot be currently handled by a study group. » FG Chair and SG17 Chair will resolve on a case by case basis.
3. Objectives
According to its scope (clause 2), the Focus Group aims to achieve the following objectives:
3.1 Terminology and definitions
To study and harmonize concepts, terminology and working definitions relevant to interoperable trust management and digital identity for humans and agentic AI.
3.2 (Pre-)standardization roadmap
To identify and study enabling technologies, trends and standardization gaps relevant to trust management, digital identity and interoperability for humans and agentic AI, in order to develop a pre-standardization roadmap.
3.3 Use cases
To identify use cases to guide future development of interoperable trust and identity solutions for humans and agentic AI across various service domains, with a view to facilitating industrial adoption.
3.4 Security assessment criteria and benchmarks
To identify criteria and benchmarks for assessing interoperability, trustworthiness, and other design characteristics (e.g. security, privacy, safety, resilience, accountability, human oversight and traceability) of human and agentic AI trust and digital identity systems.
3.5 Global collaboration platform
To provide a platform for facilitating community engagement and collaboration among relevant ITU-T Study Groups (SGs), external SDOs, industrial entities, open-source communities and academic institutions. The goal is to share best practices and disseminate knowledge for dialogue on technical-policy and potential regulatory implications of trust management and digital identity, in order to accelerate consensus on technical approaches.
4. Specific tasks and deliverables
According to its scope (clause 2), the Focus Group aims to develop the following deliverables:
4.1 Use cases and requirements analysis
Reports analysing selected use cases across sectors, identifying associated technical requirements (and references to policy, regulatory and business requirements where strictly needed for technical decisions) for interoperable digital identity and trust management for humans and agentic AI.
4.2 Architectures for identity, trust, agent discovery and interoperability
A set of architectural approaches covering identity stacks for agentic AI, cross-border trust management for human digital identity, and runtime trust control mechanisms, including agent discovery, trust establishment and interoperability across heterogeneous environments.
4.3 Trust framework(s) and lifecycle management
Framework(s) defining how trust is established, evaluated, maintained and revoked in human and agentic AI ecosystems, including trust models, lifecycle considerations, human oversight and behavioural trust signals.
4.4 Technical-policy and machine-readable trust metadata
A comparative analysis of trust frameworks and digital identity policies, including methods and specifications to represent and exchange technical-policy and trust-related information in machine-readable formats, in order to enable automated trust decisions.
4.5 Guidelines, metrics and standardization recommendations
A set of guidelines, assessment criteria and metrics for evaluating trustworthiness and interoperability, together with recommendations for future standardization activities and alignment with existing initiatives.
4.6 Coordination activities
Tasks to support liaison with relevant ITU-T SGs and FGs, industrial entities and external SDOs, in particular: collaborate with related ITU-T study groups and other standards bodies working on agentic AI, digital identity and trust management; collaborate with academia and industry to promote interoperability and accelerate the adoption of standardized approaches; organize webinars and workshops in coordination with major ITU events to collect knowledge and experience and to foster innovation. Upon completion of its lifetime, the Focus Group shall provide a final report, including the complete set of deliverables, to its parent group SG17. Annex A provides examples of expanded and specific deliverables that could be proposed as contributions to the Focus Group.
5. Relationships
The Focus Group will work through appropriate representations at meetings with ITU-T SG17. Its pre-standardization work serves to support, complement and inform the standardization activities of SG17, without duplicating ongoing efforts. Furthermore, it will liaise with other relevant ITU-T SGs and FGs as necessary to ensure coherence, promote synergies and avoid duplication. The Focus Group will collaborate with other relevant entities and groups in accordance with Recommendation ITU-T A.7. These include governments, inter-governmental entities (in particular the ITU GCC), non-governmental organizations (NGOs), policy makers, SDOs, industry forums and consortia, companies, academic institutions, research institutions and other relevant organizations dealing with aspects of trust management and digital identity for agentic AI from their own perspectives. The Focus Group will organize webinars and workshops to promote the FG activities, encouraging both ITU members and non-ITU members to jointly contribute to this Focus Group and its objectives.
6. Structure
The Focus Group may establish a sub-group structure if needed.
7. Parent group
The parent group of the Focus Group is ITU-T SG17 (Security) and relevant Questions.
8. Leadership
See clause 2.3 of Recommendation ITU-T A.7.
9. Participation
See clause 3.1 of Recommendation ITU-T A.7. A list of participants will be maintained for reference purposes and reported to the parent group.
10. Administrative support
See clause 5 of Recommendation ITU-T A.7.
11. FG financing
See clause 4 of Recommendation ITU-T A.7.
12. Meetings
The Focus Group will conduct regular meetings. The frequency and locations of meetings will be determined by the Focus Group management. The Focus Group will use remote collaboration tools to the maximum extent possible. Meeting dates will be announced by electronic means (e.g. e-mail and website) at least four weeks in advance.
13. Deliverables
See clause 8 of Recommendation ITU-T A.7.
14. Working language
The working language is English.
15. Approval of deliverables
Approval shall be obtained by consensus, in accordance with clause 8.2 of Recommendation ITU-T A.7.
16. Working guidelines
Working procedures shall follow the procedures of Rapporteur meetings (Recommendation ITU-T A.1). No additional working guidelines are defined.
17. Progress reports
See clause 3.6 of Recommendation ITU-T A.7.
18. Announcement of Focus Group formation
The formation of the Focus Group will be announced via TSB Circular to all ITU membership, via the ITU-T newslog, press releases and other means, including communication with the other organizations involved.
19. Milestones and duration of the Focus Group
The Focus Group lifetime is aligned with clause 2.2 of A.7.
20. Intellectual Property Rights
See clause 7 of Recommendation ITU-T A.7.
Annex A — Examples of expanded and specific deliverables
Annex A provides examples of expanded and specific deliverables that could be proposed as contributions to the Focus Group.
A.1 Agentic AI digital identity
A.1.1 Design of an identity stack for agentic AI
The aim is to define a coherent identity stack for agentic AI that distinguishes and articulates the different layers of identity involved in agentic systems, such as attestation, identification, authentication, authorization, delegation and agent discovery. This work should clarify how these layers relate to each other in distributed agentic environments.
A.1.2 Analysis of the properties of a delegation artefact
The aim is to analyse the core properties that a delegation artefact should have in the context of agentic AI, including how it relates to existing authorization and identity standards. This work helps identify the characteristics of a robust delegation mechanism for machine-mediated action.
A.1.3 Analysis of requirements for agentic identity, authentication, authorization, delegation and agent discovery in payment use cases
The aim is to understand how agentic AI can operate in payment contexts while meeting the requirements for identity, authentication, authorization, delegation and agent discovery. This includes an analysis of how agentic identity models fit with existing and emerging payment initiatives, including intent-based payment models and industry efforts from actors such as Mastercard and Visa.
A.1.4 Evolution of X.509 to address agentic AI requirements
The aim is to study whether and how X.509-based approaches can evolve to support the requirements introduced by agentic AI. To note that X509 is one of multiple mechanisms for trust management of digital identity.
A.1.5 Analysis of logical languages for access policy and their limitations
The aim is to study logical languages that can be used to express access control and delegation policies in agentic environments and to assess their suitability, expressiveness, interoperability and operational limitations.
A.1.6 Study on new conceptual layers adapted to agentic requirements
The aim is to explore whether the traditional abstraction layers used in networking and distributed systems are sufficient for agentic AI, or whether additional conceptual layers are needed. This work should provide a structured way to reason about agentic systems beyond protocol-level interactions and help frame future standardization efforts.
A.1.7 Knowledge graph on the agentic AI domain
The aim is to build and maintain a knowledge graph that captures the literature review in the agentic AI domain. This is intended to support ongoing research, identify relationships across fragmented workstreams and ensure that technical and policy discussions remain anchored to the evolving state of the art.
A.1.8 Security gap analysis of current open-source agents
The aim is to analyse current open-source agent frameworks and implementations in order to identify security gaps related to identity, authentication, authorization, delegation and agent discovery. This includes understanding how these agents are identified, how they authenticate to services, how permissions are managed, how actions are delegated and where risks emerge in practice. The purpose is to ground the work in the reality of deployed and emerging open-source ecosystems.
A.2 Agentic AI trust framework
A.2.1 A reference framework for agentic AI trust management
A layered reference framework for agentic AI trust management. It defines the relationship between technical trust mechanisms (such as cryptographic attestation, behavioural monitoring and capability declarations) and governance structures (such as technical-policy enforcement, oversight and escalation pathways). The framework will inform architectural design elements and interoperability specifications for deployment contexts involving both human and agentic AI digital identity workstreams.
A.2.2 A trust lifecycle framework for agentic AI
A study and specification of trust as a dynamic and time-bound property, covering initial trust establishment, continuous trust evaluation, trust degradation and revocation, including the conditions and evidence criteria required at each lifecycle stage for both agent-to-agent and human-to-agent interaction contexts.
A.2.3 A taxonomy of trust models and their applicability to agentic AI
A comparative analysis of existing trust models, including zero-trust architectures (as applied to dynamic authorization rather than network-perimeter contexts), federated trust, delegated trust and risk-adaptive trust, and their fitness for agentic AI contexts, identifying required extensions or new model constructs to address the specific characteristics of autonomous, goal-oriented agents operating in dynamic environments.
A.2.4 A framework for human oversight integration in agentic AI trust systems
A study on the mechanisms by which human oversight is embedded in trust architectures, including intervention triggers, escalation criteria, audit-trail requirements and accountability assignment, ensuring that trust frameworks preserve meaningful human control without becoming operationally prohibitive at scale.
A.2.5 Trust interoperability requirements across jurisdictions and sectors
An analysis of the regulatory, technical and governance conditions required for trust frameworks to interoperate across national and regional jurisdictions and sector-specific deployments, including financial services, healthcare and critical infrastructure, together with a gap analysis against existing ITU-T and relevant external standards.
A.2.7 A reference architecture for the agentic AI trust control plane
An architectural specification of the components and protocols required to enforce trust decisions at runtime, including technical-policy decision points, technical-policy enforcement points, trust signal propagation across agent hierarchies, and the coordination mechanisms required to maintain consistent trust state in distributed and multi-agent environments.
A.2.8 A framework and data model for behavioural trust signals in agentic AI
A study on how observed agent behaviour contributes to continuous trust evaluation, including the definition of behavioural baselines, anomaly and deviation detection, and intent inference from action sequences, producing a taxonomy of behavioural signal types and a proposed data model for representing and communicating behavioural evidence within the trust lifecycle framework.
A.3 Human digital identity
A.3.1 Study of selected use cases
The aim is to understand selected use cases that will anchor the Focus Group work to reality. The selected use cases include education, displaced persons, financial services and health.
The deliverable is: a report containing a set of legal, business and technical requirements per use case.
A.3.2 Architecture to scale trust management across borders while respecting sovereign implementations
The aim is to ensure that identity data can be safely exchanged and interpreted between different domestic implementations, while embracing a diverse range of sovereign technical and architectural decisions informed by local context (e.g. values, language, challenges).
The deliverables are: an analysis of existing conceptual models and deployed implementations for cross-border agent discovery and trust establishment, including a review of approaches adopted in operational frameworks (e.g. ICAO, AAMVA, AustRoads, eReg), with the objective of identifying common patterns, assumptions and limitations; a reference architecture to support scalable and interoperable trust establishment across jurisdictions, enabling cross-border digital identity interactions while respecting sovereign policies, legal frameworks and sector-specific requirements.
A.3.3 Comparative analysis of digital identity policies and their translation in a global machine-readable format
The aim is to analyse existing policies for digital identity across the globe and develop a mechanism to translate and communicate them when a transaction takes place between two different domestic implementations. Information includes minimum requirements for trust in a transaction, how trust has been achieved, and how actors can trust each other.
The deliverables are: a tool to compare current digital identity policies; a technical specification to transform technical-policy into machine-readable format.
A.3.4 Standardized set of rules and guidelines for selected sectors (e.g. education)
Countries and ecosystems adopt rulebooks to enable domestic and cross-border interoperability of credentials.
The deliverables are: a standardized rulebook for sector-specific credentials (e.g. diploma); filling gaps in emerging rulebooks for champion use cases (e.g. mDL for bank account opening).
A.3.5 Study on security approaches for global interoperable identity wallets beyond what is ready to be standardized
Countries should understand potential gaps between security requirements and user outreach and inclusivity. The end goal is to harmonize existing wallet security requirements in jurisdictions that need it.
The deliverables are: a structured list of security solutions for identity wallets (e.g. HSM, eSE) with their limitations; a gap analysis of interoperability areas to be addressed (e.g. attestations for platform secure elements); a study on device penetration in selected areas of LDCs, including a set of guidelines for achievable levels of security.
A.3.6 Study on the relevance of digital identity to national development indicators
In order to ensure that the Focus Group is data-driven, policymakers leverage statistical analysis for data-driven decision making.
The deliverables are: a statistical analysis of the correlation between identity and development factors (e.g. levels of education, healthcare access, etc.); a data-visualization tool.