By Dražen Kapusta and Terence Tse

As AI agents increasingly power cross-border commerce, a critical gap is emerging: digital infrastructure can exchange data but not understand it. Without a semantic interoperability “trust layer” — a shared, machine-readable framework for verifying what actually occurred in a transaction — AI-driven economies risk compounding errors, governance failures, and systemic opacity at unprecedented speed.

If someone is quietly congratulating themselves on finally understanding digital identity and AI compliance, here’s a reality check. Building trustworthy infrastructure involves more than just wallets and regulations. The past two years have seen the development of infrastructure like the EU Digital Identity Wallet, the AI Act, and the Digital Euro. However, the landscape has shifted. A new question is emerging, one less about infrastructure components and more about whether those components can understand each other. This is about semantic interoperability – the ability for machines to not only exchange data but also interpret its meaning across borders, systems, and languages. Unlike previous technical debates, this issue could determine whether our digital future truly succeeds. The reason for this new – and perhaps inevitable – development is simple: around the world, the foundational pieces of an autonomous economy are being assembled. But they were all put together without the layer that allows them to be understood by one another.

Many Conversations, One Missing Layer

We don’t need to look far to see the lack of semantic interoperability. In Brussels, technical experts warn that the EU Digital Identity Wallet – Europe’s leading digital infrastructure – is being developed with a critical flaw. The draft Implementing Regulations support basic data exchange but omit the semantic layer necessary for machines to comprehend what they’re reading. In other words, a professional qualification issued in Spain cannot be automatically understood by a German system. An educational credential cannot carry meaning across institutions. A regulatory attestation cannot be verified across borders without human intervention. In short, Europe is constructing an identity infrastructure that enables machines to read but not understand.

AI leaders from around the world convened at the India AI Impact Summit 2026. Sam Altman of OpenAI called for a global AI regulatory framework similar to the International Atomic Energy Agency. French President Emmanuel Macron proclaimed Europe a “safe space” for regulated AI. UN Secretary General António Guterres warned that no child should be a test subject for unregulated AI.

Innovation and governance, they all agreed, must go hand in hand.

Yet among these conversations lies a gap. Digital infrastructure is being built without interoperability. Can these setups exchange data or apply the same governance principles? The policymakers governing AI are focused on principles – should AI be regulated? No one seems to be creating the layer that links these individual, self-reliant systems.

The Problem

Sooner or later, issues caused by such divergent development pathways without interoperability will catch up with us. Imagine a routine cross-border transaction. A Spanish supplier’s AI agent negotiates with a German buyer’s agent. The supplier ships goods. The carrier logs delivery. The buyer’s system triggers payment. But the supplier’s product identifier uses a different schema than the buyer’s ERP. The carrier’s delivery event employs different semantics than the buyer’s proof-of-delivery requirements. The AI agents, acting rapidly, lack a shared layer of meaning.

Within minutes, three systems claim incompatible realities. The buyer’s agent disputes payment. The supplier’s agent raises a breach. Treasury automatically suspends the vendor. By the time a human investigates, dependent actions have already been carried out: reorder loops, penalty clauses, and credit holds. This isn’t very different from a situation where three witnesses to the same car accident, each speaking a different language and following different legal systems, submit their reports. The accident happened. Everyone agrees something took place. But without a shared framework to interpret what each report means, the insurance claim collapses — and by the time a translator arrives, the car has already been towed, the claim rejected, and the policy cancelled.

An Old Idea, Whose Time Has Come Again

In 1458, a merchant from Ragusa in southeastern Sicily, Italy, named Benedetto Cotrugli authored the world’s earliest known treatise on double-entry bookkeeping, thirty-six years before Luca Pacioli received most of the historical credit. Yet, what Cotrugli recognised was not merely an accounting technique. He understood that commerce on a large scale requires a shared framework of truth: a method for parties who have never met, trading across borders they cannot physically cross, to establish a mutually understandable record of what they agreed to and what they owe each other. His ledger was not just a business record. It was a social contract.

The challenge we face today remains largely the same. We operate in a global marketplace — an AI-driven economy enabling an increasing number of cross-border and cross-system transactions at unprecedented speeds. What we lack is the equivalent of Cotrugli’s ledger: not a record maintained by one party or another, but a shared, verifiable, collectively authoritative account of what truly transpired. The difference is that this time, the ledger cannot be paper-based, nor can it wait until the month-end close. It must be machine-readable, policy-aware, and capable of operating at the same speed as the transactions it manages.

The Missing Layer

Much of today’s policy discussion remains a level too shallow. Governments debate AI governance. Companies discuss technological expertise. Standards organisations debate formats and schemas. But governance without machine-speed evidentiary infrastructure is fragile, and expertise without semantics merely accelerates misunderstanding. A machine can process data without understanding it. Two systems can exchange records without agreeing on what has occurred. In a zero-trust environment, that is not resilience. It is vulnerability disguised as automation and speed.

This is why the future’s AI-powered NEO World will need a trust layer.

Such a layer must do more than just store records. It must ensure that when a meaningful event takes place between parties, that event produces a shared, verifiable, co-attested object that neither party can unilaterally alter afterwards. It must carry evidence, not merely reference it. It needs to be machine-readable and machine-interpretable. Additionally, it must allow authorised third parties – auditors, regulators, counterparties, courts, or AI agents themselves – to determine later, with high confidence, what happened and under which conditions.

This is the logic behind what we call NEO accounting. This is important for three reasons that extend far beyond accounting. Firstly, it means that the governance rules that apply to a transaction are embedded within the transaction record at the time of signing, not referenced from an external document that may later change. Secondly, it means that AI agents can be assigned cryptographically bounded mandates — a digital fence that specifies exactly what they are authorised to do and automatically triggers a downgrade if they breach it — without requiring human approval for each individual action. Thirdly, it means that every participant in a network builds a trust score based not on what they claim about themselves, but on what their transaction history demonstrates.

Cotrugli understood this last point intuitively, five centuries before it became an engineering problem. The merchant’s reputation, in his telling, was not a title or a credential. It was what the ledger proved, transaction by transaction, year by year.

What This Means for the Future

For business leaders, the concern is operational risk. As AI agents become standard participants in procurement, logistics, and financial transactions, the gap between what different systems believe to be true will widen faster than any manual audit can monitor. Organisations without a shared truth infrastructure will face compound error cascades at a speed that makes recovery exponentially harder.

For regulators, the message is equally clear. The EU AI Act, the Digital Identity Wallet, and AI governance frameworks from New Delhi all assume AI decisions will be traceable and auditable. But traceability requires capturing the trace at the moment of action — not reconstructing it from logs neither party independently trusts. Policy without the technical infrastructure to enforce it is not governance. It is simply an aspiration.

For architects of the Digital Single Market, the stakes are highest. Europe’s investment in digital identity will only realise its potential if credentials carry machine-interpretable meaning across borders — if a Spanish qualification is not merely displayed to a German system but understood by it. Without semantic interoperability, Europe risks building something that appears unified but functions as a collection of fundamentally opaque national systems.

The standards embedded in upcoming AI and digital frameworks will shape commerce for a generation. The window to include a semantic interoperability layer is closing — once specifications are finalised, extending them becomes a highly complex political and technical challenge. This is not a future problem. It is a present decision with long-term consequences.

Cotrugli wrote not only for merchants but also for the architecture of commerce as a whole. He understood that the system of shared truth he described was a vital choice: not just a tool for efficiency, but a decision about what kind of economy — and what kind of trust between strangers — a society wished to build. We face the same choice today as in Cotrugli’s time, only at a much greater scale and with things running at a speed he could not have imagined.

About the Authors

Dražen Kapusta is the founder of COTRUGLI Business School and HashNET. He leads the COTRUGLI initiatives, focusing on AI-augmented Vanguard leadership, NEO Finance, blockchain, SDGs, and digital sovereignty. Dražen advises UN and EU bodies on AI and blockchain strategies.

Terence Tse is Professor of Finance at Hult International Business School and co-founder at the AI Native Foundation. He is also co-founder and Executive Director of Nexus FrontierTech.