Why AI Agents Need Verifiable Organizational Identity

As the deployment of AI agents accelerates across the global digital economy, the question is not what these agents can do, but how the organizations behind them can be identified and held accountable.

---

---

We have entered the ‘agentic era’. AI agents now interact with the internet, software, and each other, often acting autonomously on behalf of the people and organizations that deploy them.

This is upending trust models that assume a human is “in the loop” and ultimately accountable. When an agent commits an organization to an obligation, such as booking a flight or transferring money, the interaction appears ordinary from the outside, while the accountability behind it remains unclear.

The accountability challenge stems from the difficulty of reliably establishing which organization stands behind the agent, who within that organization authorized it to act, and whether the action fell within the scope of authority granted to the agent. With automated traffic accounting for 53% of all web activity in 2025, and AI agents emerging as a distinct category alongside conventional bots, solving this challenge will soon become a precondition for trusting any interaction at all.

Connecting machine identity to organizational accountability

The good news is that progress is being made. Ongoing research into techniques to reliably identify and track the AI models that agents are built on, even as the underlying neural networks change and evolve over time, is showing promise. Yet these “model identifiers” do not address who is accountable for an agent and its actions. For accountability to hold, an agent’s identity must connect upward to the organization behind it and the people authorized to act on its behalf.

The Legal Entity Identifier (LEI) already provides this organizational anchor. It is a globally recognized, unique identifier for legal entities that is already embedded in regulatory reporting across more than 100 jurisdictions. Its digital counterpart, the verifiable LEI (vLEI), extends that identity into digital interactions, allowing counterparties to computationally verify the identity, authority, and role of a person or a system acting on behalf of a legal entity. The LEI can also serve as the organizational anchor for other digital solutions that support the same objectives, enabling identity interoperability among them.

How verified organizational identity can work in practice

As organizations increasingly deploy AI agents to interact directly with customers via chat and voice calls, there is an urgent need for customers to verify that an agent genuinely represents the organization it claims to represent.

GLEIF’s Global vLEI Hackathon, which drew more than 110 submissions from teams worldwide, showed how this could work. Agent-Anchor, from the team YourData, was named runner-up in the Industry 4.0 category in December 2025. It demonstrated how assigning vLEI identities to a company’s AI agents enables a customer interacting with a voice or chat agent to verify the agent's origin in real time during a call or other interaction.

While Agent-Anchor is a prototype, not a deployed product, it demonstrates the potential of verifiable organizational credentials, bound to an automated agent, that can be checked at the moment of interaction.

A complementary example comes from Curvy Protocol, whose approach to identity-anchored confidential payments can also extend to AI agents. Curvy demonstrates how an agent can be given a payment identity with defined counterparty rules and appropriate controls. GLEIF’s recent work explores how approaches such as this could be anchored to vLEI-based organizational identity and delegated authority, creating a verifiable link between an AI agent, the legal entity it represents, and the authority under which it acts.

Turning autonomous agents into accountable agents

Though AI agent adoption is accelerating, without a clear link between an agent and an accountable individual or organization, the risks outweigh the benefits and scalable adoption will be difficult for both enterprises and regulators to support.

Verifiable organizational identity is the layer that changes that. It makes agents accountable by connecting what a system does back to the legal entity responsible for it and the authority under which it acted. That is the difference between an agent that merely operates and an agent that can be trusted to operate on your behalf.

For organizations beginning to deploy agents, the practical implication is to treat identity and authority as part of the deployment, not as something to retrofit later. The question to ask of any agent acting in your name is not only whether it works, but whether a counterparty can verify which entity stands behind it and what it was authorized to do.

The organizational identity infrastructure to support that already exists in the LEI and vLEI. The model-identity research that would complement it is advancing, but is not yet in production, which is why the organizational layer should be put in place now.

Strengthening digital trust in the AI era

The intersection of AI, digital identity, and digital security was the focus of my Trust Talks conversation with Professor Damian Borth, Chair of Artificial Intelligence and Machine Learning at the University of St. Gallen and co-founder of Weight Space Labs. We discussed how AI is reshaping digital security, why binding identity to a changing neural network is so difficult, how AI can be used to check AI without becoming its own judge, and why the identity of the organization behind a system remains the fixed point when the technology around it keeps moving.

Listen to the full Trust Talks episode to explore what verifiable organizational identity means in a world where software increasingly acts on our behalf. Trust Talks is available across YouTube, Spotify, and Apple Podcasts: linktr.ee/TrustTalks

---