How YourData Is Exploring Execution Trust for AI Agents with the vLEI

Jaegyeong Yeom, co-founder of YourData, explains how its Agent-Anchor prototype is designed to use the verifiable LEI (vLEI) to connect AI-generated actions to verified organizational identity and roles, ensuring executed actions are authorized before reaching business systems.

---

---

AI agents increasingly act for companies in interactions that previously involved a person. What first motivated YourData to explore the verifiable LEI (vLEI) as a trust foundation for AI agents?

Our starting point was how an external party could verify that an AI was acting on behalf of an organization.

In traditional enterprise environments, an employee’s account, internal identity and access management (IAM) controls, approval processes, and contractual relationships often establish who is acting and with what authority. AI agents can now communicate with external parties, generate documents, use tools and application programming interfaces (APIs), and initiate actions with real business consequences. This makes it essential to verify which legal entity stands behind an agent, the organizational capacity in which it is acting, and the authority supporting its actions.

The vLEI offered a relevant foundation for this work. Key Event Receipt Infrastructure (KERI) supports verification of identifier control, changes in control, and delegation, while the vLEI connects this cryptographic foundation to verified legal entity identity and organizational roles. Together, they provide a basis for linking verified organizational identity and authority to AI-mediated execution.

Your Agent-Anchor solution was named runner-up in the Industry 4.0 category of GLEIF’s Global vLEI Hackathon 2025. How is it designed to enable a counterparty interacting with an AI callbot, chatbot, or automated message to computationally verify which organization stands behind it and under what authority?

Agent-Anchor begins with a simple observation: verifying a phone number, domain, communication channel, or even a voice does not establish that an AI is acting for the organization it claims to represent.

The key question therefore changes from whether a message looks or sounds authentic to whether the recipient can verify the legal entity behind it, the organizational capacity involved, and the authority supporting the specific action. This distinction matters because authenticating an AI or its communication channel does not make every action proposed through that channel an official act of the organization.

Using the vLEI, Agent-Anchor enables a counterparty to computationally verify the legal entity and relevant organizational capacity behind an interaction, rather than relying on the AI’s own claims. It then connects that organizational identity to the proposed action and assesses whether the request falls within the applicable mandate, purpose, scope, and conditions of authority.

YourData distinguishes between verifying an agent’s organizational identity and authorizing a specific action. Why does that distinction matter, and why does the vLEI credential attach to the legal entity and organizational role rather than to the AI system itself?

The vLEI is anchored in the legal entity, and vLEI role credentials can verify the organizational roles of people acting on its behalf. As these credentials do not attach to the AI system itself, they provide a common institutional reference across organizations, platforms, and jurisdictions, while allowing each organization to retain its own internal controls and policies.

Agent-Anchor therefore distinguishes between the agent, which proposes or performs an action, and the actor, the person or legal entity to which that action is organizationally attributable. An AI’s technical ability to generate or perform an action does not make the AI the ultimate institutional reference point.

Verifying representation is only the first step. A connection to an organization and role does not mean that every action is authorized. For Agent-Anchor to permit an action, the action must also satisfy the relevant mandate, scope, purpose, policies, approval requirements, and execution conditions. The vLEI provides the verifiable organizational trust foundation, while Agent-Anchor connects that foundation to the authority required for a specific action.

How can verifiable organizational trust translate into action-specific authority and execution?

The Agent-Anchor Executor operates at the point where an AI-generated proposal may become an action with real organizational effect. For each proposal, it evaluates who the action is attributable to, which role or mandate supports it, whether the authority and delegation remain valid, and whether the target, purpose, scope, policies, approvals, limits, and execution conditions are satisfied. It also checks that the action ultimately executed matches what was authorized.

This addresses what we call the execution-trust gap. Verifying an organization and role establishes a trust foundation, but it does not determine whether a particular action may proceed. The Executor applies that authority boundary before the proposal is reflected in a business system or external environment, connecting the basis for authorization and the execution outcome to verifiable evidence.

The model is designed for use across domains, including payments, trade, and supply chains, with policies and integrations adapted to each environment. YourData is currently validating and extending the model through proof-of-concept and integration work with enterprise partners.

Can you walk us through one agentic transaction from start to finish? Which checks happen at which point, and what does the counterparty see?

Consider a hypothetical example of an AI agent negotiating a cross-border purchase for a company. Before it proposes a commercial commitment, the supplier should be able to verify the legal entity and organizational capacity behind the proposed action. Agent-Anchor would then assess whether the commitment falls within the relevant mandate and approval conditions.

If the transaction moves to payment or financing, a new authority check is required. Permission to negotiate commercial terms does not automatically authorize issuing a financial instruction. The relevant mandate, transaction limits, and approvals must be assessed, while financial institutions continue to apply their own compliance, risk, and transaction controls. The same principle applies to trade documents, logistics instructions, and customs submissions, each of which may involve different roles and rules.

The counterparty does not need access to the company’s full internal IAM or approval process. It needs sufficient verifiable evidence to establish which legal entity and organizational capacity stand behind the action, whether the required authority conditions were met, and whether the executed outcome matches what was authorized. The vLEI provides the cross-boundary trust reference, while Agent-Anchor connects it to each action’s specific authority and evidence.

You have said Agent-Anchor is designed to preserve the autonomy of national identity systems, enterprise controls, and sector-specific frameworks. How do you see verifiable organizational identity working alongside infrastructures such as enterprise IAM and the European Business Wallet as those frameworks develop?

Verifiable organizational identity should complement existing infrastructure, not replace it. Enterprise IAM governs accounts, roles, privileges, and access within an organization. Initiatives such as the European Commission’s proposed European Business Wallets are intended to help businesses identify themselves, present credentials, and interact across organizational and jurisdictional boundaries. These systems operate at different levels and need not converge on one technical model.

The vLEI provides a separate institutional reference point for presenting and verifying legal entity identity and organizational roles across organizations, platforms, and jurisdictions. KERI supplies the cryptographic foundation for verifiable control, continuity, and delegation.

Agent-Anchor is neither a wallet nor an enterprise IAM system. It operates at the execution-trust layer, connecting verifiable organizational trust to the mandate, policies, approvals, and conditions governing a particular action. The underlying environment may continue to use enterprise IAM, wallet credentials, or sector-specific controls. Our design principle is interoperability while preserving existing institutional models. Each environment retains its governance and controls, while organizational identity and action-specific authority remain verifiable across boundaries.

How does Agent-Anchor address scenarios such as revoking or changing an agent’s authority mid-interaction, verification by parties outside the ecosystem, or determining legal liability for an agent's decisions?

Agent-Anchor addresses a defined part of the trust problem. However, it does not replace an organization’s operational rules, the wider trust ecosystem, or applicable law.

The model can represent revocation. KERI supports verification of changes in control and delegation, while the vLEI Ecosystem Governance Framework provides mechanisms for issuing, maintaining, and revoking organizational credentials. Within an enterprise, Agent-Anchor or connected identity and authorization systems can enforce changes to an agent's authority.

External verification is more complex because the wider vLEI and vLEI Issuer (QVI) ecosystem must promptly access and consume current trust and authority information. If authority changes during a transaction, organizational or sector-specific policies must determine whether the action stops, continues, or requires renewed approval.

The model can provide evidence of which legal entity and organizational capacity stood behind an action, what authority supported it, and what was executed. It does not determine legal liability, which remains governed by law, regulation, contracts, and institutional governance. Nor does it address model alignment, hallucinations, training, or every aspect of AI safety. It focuses on the point at which AI-generated intent becomes an official organizational action with external consequences.

Looking ahead, what new use cases could emerge from applying Agent-Anchor’s execution trust model to areas such as cross-border payments and digital trade documents, and what has to be in place first?

The broader opportunity is to move from verifying the origin of information to verifying the authority behind an action.

In cross-border payments, an AI or enterprise system may generate a payment instruction, but the receiving system must still confirm it links to the correct legal entity, organizational role, mandate, recipient, purpose, transaction limits, and approvals. The same principle applies to digital trade documents. For invoices, electronic bills of lading, customs submissions, and transfers of rights, parties need to verify who was authorized to create, approve, modify, submit, or act on the document, and whether the resulting action remained within scope. Similar use cases could emerge for Digital Product Passports, corporate wallets, and digital assets.

Deployment at scale requires several foundations to mature together: verifiable legal entity identity and organizational roles, reliable delegation and current authority information, integration with enterprise and sector-specific policies, interfaces to execution systems, interoperable evidence, and effective governance. KERI provides a foundation for identifier control and delegation, while the vLEI adds verified legal entity identity and organizational roles. Agent-Anchor connects those foundations to action-specific authority and execution. This is what we mean by execution trust.

---