When Machines Act, Trust Must Be Verifiable

From digital identity to accountable agency. Dr. Pramod Varma, Co-founder & Chief Architect, Networks for Humanity (NFH), explains why the next phase of Digital Public Infrastructure must connect organizations, people, and the authority they delegate.

---

---

For much of economic history, trust traveled with people. A signature, a seal, or a personal introduction linked a business commitment to someone accountable. Digital networks allowed those commitments to cross borders in seconds. Artificial intelligence is now changing who initiates them.

As software begins to negotiate, purchase, and instruct other systems on our behalf, a familiar institutional requirement takes on new urgency: every consequential action needs a verifiable basis of authority.

My work on India’s digital public infrastructure, and now on open networks at Networks for Humanity, has reinforced a practical conviction. Participation expands when people and organizations can establish trust through shared infrastructure. The AI era makes that infrastructure more necessary, and places new demands upon it.

The past: establishing identity at scale

The internet made information widely accessible, but businesses still faced a basic difficulty: identifying who was on the other side of an interaction. Company names could be ambiguous. Records were dispersed across jurisdictions. Companies often had to repeatedly collect and verify documents.

The consequences extended beyond administrative inconvenience. After the global financial crisis, consistent legal-entity identification became an international priority. In 2011, the G20 asked the Financial Stability Board to develop recommendations for a global Legal Entity Identifier and the governance that supports it. The resulting Global LEI System provided a common reference for identifying organizations across markets. Its public value rests on both standardized identity data and the institutions responsible for maintaining it.

India’s experience offers a complementary lesson. Aadhaar, UPI, and DigiLocker illustrate how shared capabilities in identity, payments, and digital documents can support services built by many institutions. In helping to architect these systems, I have come to regard identity as a means of enabling participation. Its value becomes clear when a person or business can use it to access services and transact with greater confidence.

Aadhaar, Jan Dhan bank accounts, and mobile connectivity helped adult account ownership rise from 35% in 2011 to 80% in 2017. UPI made low-value digital payments practical for millions of merchants, while direct benefit transfers and Aadhaar-enabled banking brought financial services closer to households. For pensioners, digital life certificates reduced dependence on physical visits, demonstrating how verifiable identity can improve access and preserve dignity — see World Bank: Identification and Financial Inclusion.

The same principle applies to small businesses. India’s Udyam registration system uses Aadhaar and enterprise information to simplify formal recognition; together with Udyam Assist, it recorded more than 89 million MSME registrations by July 2026. Udyam Assist extends recognition to informal micro enterprises, and its certificates are accepted for access to priority sector lending benefits, creating a practical route into formal financial services. The lesson is clear: identity adoption becomes more compelling when it opens access to services that businesses value.

This is the wider promise of digital public infrastructure: essential capabilities that can be reused across many applications, supported by clear governance and open standards. A common foundation allows innovation to spread beyond the institution that first builds it.

Organizational identity belongs within that foundation. GLEIF’s recent article, The Missing Layer, argues for the Global LEI System as a global Digital Public Infrastructure (DPI) for business identity. National registries and identity systems remain essential; a shared global identifier can help connect them across borders.

The present: making trust digitally verifiable

AI increases the distance between an action and the person who would previously have performed it. An agent may respond to a customer, evaluate a supplier, or prepare a payment instruction. Its fluency tells the recipient little about the institution behind it or the authority under which it operates. As convincing text and voices become easier to produce, recipients need evidence they can verify independently of the content itself.

Three questions therefore need distinct answers. Which legal entity is represented? Who has authorized the activity? What, precisely, may the agent do?

GLEIF’s recent work on AI agents places organizational identity at the center of this accountability problem. Identifying an AI model or authenticating a software endpoint can be useful. Establishing the organization behind an interaction provides a different, necessary piece of evidence.

The verifiable Legal Entity Identifier, or vLEI, puts organizational identity in a form software can verify. It connects an organization’s LEI with cryptographically verifiable credentials. Associated role credentials establish the identity and organizational role of people acting for that entity. A recipient can verify these claims through a trust chain rooted in GLEIF, rather than relying on a document's appearance or a message's assertions. A credential is useful because it carries a claim, identifies the issuer, and provides a way to check its authenticity and integrity. Its reliability also depends on the rules governing issuance, maintenance, and revocation. The vLEI Ecosystem Governance Framework addresses these institutional responsibilities by leveraging cryptography.

There is an important boundary here. Proof that a person holds an organizational role does not, by itself, authorize every action proposed by that person’s AI agent. The mandate, its limits, and the controls governing execution must also be established. GLEIF’s September discussion of YourData’s Agent-Anchor prototype makes this distinction explicit: vLEI provides the organizational foundation, while additional mechanisms connect it to permission for a particular action — see Exploring Execution Trust for AI Agents.

This separation should guide implementation. Verifiable identity establishes who stands behind an interaction. A mandate defines permitted activity while operational controls decide whether the proposed action may proceed.

The future: delegating work while retaining control

Consider a small manufacturer using an AI agent to source components from an overseas supplier. The agent may compare quotations and negotiate delivery dates. The manufacturer may permit it to place orders within a defined budget, while reserving payment approval for a finance officer.

In a well-designed system, the supplier could verify the manufacturer’s organizational identity and receive evidence of the authority supporting the order. The mandate would identify the permitted activity, spending limit, and validity period, and be bound to the agent presenting it. A request to release payment would require its own authorization checks. If authority were withdrawn, subsequent actions would need to reflect that change.

The broader opportunity is to enable such verification across borders and networks. A business should be able to choose its service providers and still present evidence that counterparties can independently assess. Otherwise, each platform will establish its own closed circle of recognized organizations and authorized agents, recreating the fragmentation that DPI seeks to overcome.

At Networks for Humanity, our work on Beckn and the Finternet is concerned with enabling value exchange across open, interoperable networks — traditional or agentic. Credentials and trust infrastructure are integral to that ambition. One example is our work on the trust layer via the Decentralized Directory protocol (DeDi), in collaboration with the Linux Foundation (LFDT). Organizations need to carry verifiable evidence of identity and authority across services, while each service retains responsibility for its own acceptance rules and controls.

Three priorities deserve particular attention as this develops.

First, make participation practical and worthwhile for smaller organizations. We should set a global ambition to equip more than 300 million micro, small, and medium enterprises with LEIs and vLEIs, enabling them to establish verifiable organizational identity as they join the emerging agentic economy. This will require supporting formalization where needed and making credentials useful in everyday business.

Today, most small businesses lack a strong, visible incentive to obtain and maintain an LEI. We should work with banks, payment providers, major buyers, trade platforms, and public agencies to connect LEI/vLEI adoption to tangible benefits: faster onboarding, lower verification costs, better access to working capital and trade finance, and easier discovery by trusted buyers and their AI agents. India’s experience suggests that adoption follows when incentives align.

The cost and complexity of obtaining and using credentials should be dramatically reduced to ensure every business — large or small — can leverage LEI/vLEI. Simple onboarding, usable tools, and competitive provision will matter as much as technical standards. Institutions that benefit from easier verification should help fund issuance and renewal, embedding these services into existing business relationships.

Second, disclose only what the interaction requires. A counterparty needs sufficient evidence to verify identity and relevant authority. It should not need access to an organization’s entire internal hierarchy or approval process. Build privacy and commercial confidentiality into the design from the start. This is where vLEI can help, given its inherent capabilities for selective disclosure and zero-knowledge proofs.

Third, preserve control throughout the transaction. Authority changes. Employees leave, budgets are revised, and mandates expire. Verification must account for those changes at the point of action. So, creating self-managed, verifiable authorization through verifiable credentials is critical, rather than building centralized databases. Organizations also need clear arrangements for revocation, transaction stoppage, exception handling, and dispute resolution.

As a GLEIF board member, I see an opportunity to carry an established public purpose into a new setting. The LEI provides a shared reference for organizational identity. The vLEI makes that identity, and the roles of people representing an organization, digitally verifiable. Together with interoperable mandates and effective controls, they can support a world in which software acts with clearly bounded authority.

Any ID system should let people and businesses participate in the digital economy and confidently delegate useful work. As machines gain the capacity to act, our infrastructure must preserve people's and institutions' ability to decide, verify, and remain accountable. A system intended to widen access should measure its success by the opportunities it creates!

---