Agentic AI & Digital Identity: Building Trust & Accountability and Autonomous Systems

Context

AI systems are increasingly evolving from tools into autonomous or semi-autonomous agents that make decisions, initiate actions, and interact with other systems and actors. This shift changes the central governance question. It is no longer mainly how data is processed, but who is accountable when a machine acts.

Every autonomous action raises three questions that systems must be able to answer in real time:

1. Who is responsible for it?
2. What was the agent permitted to do, and on whose authority?
3. Where did value (such as money, data, obligations) flow as a result?

Answering these after the fact, through logs and legal review, no longer works when agents take action at machine speed and scale. At the same time, decision-making becomes more opaque, responsibility becomes fragmented, and trust in systems becomes at risk of eroding.

Responsibility, permissions, and value flows must therefore become attributable by the systems themselves: encoded in a machine-readable form which machines verify and humans audit. This is where digital identity becomes critical infrastructure.

Digital identity, be it of individuals, organizations, and potentially agents, emerges as the essential layer to:

• ensure verifiability and accountability
• assign responsibility
• manage access and rights

Challenge Framing

This challenge explores how agentic AI systems can be designed and governed in a way that ensures accountability, trust, and societal legitimacy.

It focuses on the intersection of:

• agentic AI systems
• digital identity and credentialing
• governance and responsibility frameworks

The goal is to move beyond abstract principles and develop practical, testable approaches that:

Allow systems to operate interoperably and without platform lock-in

Enable clear attribution of actions and decisions in automated systems

Ensure that humans and organizations retain meaningful control

Make responsibility and decision-making traceable and auditable

Key Problem Areas

Projects may address one or more of the following.

1. Identity for agents and actors

As agents begin to act on behalf of people and organizations, they need to be identifiable and accountable in their own right. How can AI agents be identified, authenticated, and authorized to act? And how can roles and delegation be modeled along the chain from human to organization to agent, so that it is always clear on whose authority an agent is acting?

2. Accountability & traceability

When an automated system makes a decision, it must be possible to understand and attribute it after the fact. How can decisions made by AI systems be traced back to their inputs, explained in understandable terms, and attributed to the responsible human or organizational entity?

3. Governance of automated systems

How can rules and constraints be embedded directly into systems and made enforceable, rather than living only in human-readable policy documents? And how can meaningful human oversight and intervention be guaranteed when systems act autonomously?

4. Verifiable data & decision pipelines

Agents are only as trustworthy as the data and processes behind their actions. How can the data used by agentic systems be made verifiable in origin, contextualized for appropriate use, and controlled in terms of usage rights?

5. Value and obligations between agents

When agents transact with other agents and systems, value and commitments move with them. How can the flow of value between automated actors be made transparent, attributable, and auditable?

6. Autonomy vs. control

The central design tension is enabling automation without surrendering human agency. How can systems be designed to allow agents to act independently while preserving meaningful human agency and institutional responsibility?

Expected Outputs

• Functional technical or socio-technical prototypes 
• Clearly defined use cases (e.g. public sector, B2B, civic tech)
• Documentation of:
  • identity models
  • governance assumptions
  • accountability mechanisms
• Where relevant, projects should aim to demonstrate:
  • Integration with governance processes (e.g. auditability, reporting)
  • Potential for real-world application

Note: We are aware that the output we are asking for may remain predominantly in prototype stage.

---

Challenge Partners

T3i Partner Network