AI got good enough to act on our behalf. What it still cannot do is prove who it acts for. Concordium is an AI infrastructure where every account has traced back to a verified identity since 2021. That is what AI agents need now, and why it belongs in the base layer rather than inside any one company.
An AI agent can read a supply contract, compare four vendors and negotiate a price. It cannot prove there is a real person behind it, or any specific attribute about that person.
You tell it to reorder the weekly shop and add a case of wine. It finds the wine, fills the basket, and stops. The retailer has to know the buyer is old enough, and the agent's options are to fetch you or to hand your passport to a supermarket that never wanted it and should not be storing it. So it waits. An order ninety seconds from done sits there until you next pick up your phone.
This is already commerce. Salesforce reports that AI and agents drove about a fifth of global orders in the 2025 holiday season, around $262 billion in sales. Gartner expects more than $15 trillion of B2B buying to run through agents by 2028, which would be about 90% of all business purchasing.
Intelligence is not the bottleneck.
Where It Stops
Trust has not caught up yet. Checkout.com found that a third of consumers expect at least a tenth of their purchases to be AI-driven within a year, while 27% say they trust no organisation at all to run an AI shopping agent and 24% say they will never delegate a purchase to one. Asked what would close that gap, the top three answers were spending caps (30%), instant revocation (29%) and easy cancellation (28%). Nobody asked for a better model.
The wine is incidental. An agent works fine until it has to represent someone or move real money. At that point it needs to prove something about a person who is not in the room: that they exist, that they authorised this, that they are old enough, that the money is theirs to spend. It cannot, so a human steps back in.
What is missing is a way for a stranger to check a claim about someone who is absent, without trusting whoever makes the claim, and without the person's details being handed over in the process.
That is a trust problem, and trust problems get solved by infrastructure.
Why That Points at a Blockchain
A blockchain is good at one narrow job, and it happens to be exactly this one.
It is a shared record that no single participant owns. Two parties who have never met, and who have no reason to trust each other, can rely on the same facts without relying on each other. A claim recorded there can be checked by anyone, at any time, without asking permission or waiting for an API to answer. It also puts the proof and the money in the same place, so an agent can settle a payment and evidence its authority in one motion.
The alternative is that the trust layer belongs to a company, and every agent is a tenant. The company decides who gets verified, what verification means, what it costs, and who stops being verified tomorrow. That holds until your agent has to prove something to a counterparty on a different provider, or the provider changes its mind, and an agentic economy running through one company's permission system is not an economy, it is an isolated platform.
So why not use Ethereum, or any chain that already exists? Identity there is real and improving, but it sits above the account rather than inside it. Each application decides what to verify and when, so a counterparty has to check case by case whether anything was ever verified.
An identity-first Layer-1 settles that once. It is the base network that keeps its own ledger and sets its own rules, including what counts as an account, and a rule written there holds for everything built on top. On Concordium the rule is that an account exists only where an identity provider has verified the person or business behind it. It is the same check, moved to the one place where nobody has to opt into it. That leaves room for agents that live elsewhere, which can keep running where they run and still resolve back to a verified owner here.
That is the whole of the job. A blockchain will not train a model, run inference or hold memory, and anyone selling it for that is selling the wrong thing. Concordium runs no models, hosts no inference and orchestrates no agents. “AI infrastructure” is a phrase that also covers GPUs and model APIs, and Concordium sits outside that half of it. The claim is narrower and it holds: this is an infrastructure that makes agents transactable with.
Where Concordium Came From
It was founded in 2018 by Lars Seier Christensen, who co-founded Saxo Bank and spent twenty years running it. Twenty years in that world leaves you with one reflex, and he has put it plainly: “You need to know who you’re dealing with.” Anonymity, in his phrase, was an absolute non-starter, for regulators and for any serious business. Asking that in 2018 is what put identity in the design rather than on a roadmap.
The people reflect the two worlds it set out to join. From science came Ueli Maurer, professor of cryptography at ETH Zurich and a Foundation Board member; Ivan Damgård, professor at Aarhus and a Concordium advisor, whose Merkle-Damgård construction underpins the SHA-2 hash family and therefore a great deal of the internet; and Torben Pryds Pedersen, who became Concordium's CTO in 2019 and whose 1991 commitment scheme underpins much of modern zero-knowledge cryptography.
From markets came Hans-Ole Jochumsen, former vice-chair of NASDAQ, Peter Klein, a former Mastercard executive vice president, and Michael Jackson, Skype's former chief operating officer.
The 2020 testnet already used zero-knowledge proofs, so compliance could be shown without personal data being revealed. Mainnet followed in June 2021.
Since that day, every account on Concordium has traced back to a verified real-world identity. Today the same rule covers businesses and the AI agents they deploy: identity checks, age verification and payments run on Concordium now, and the agent registry has linked agents on Ethereum and Solana back to verified owners since May. A protocol can add an identity layer this year. It cannot add the years in which every account already had one.
If the Design Is Right, Why Did It Not Scale Back in 2021?
Because in 2021 it answered a question most of crypto was not asking. Mandatory identity looked like friction to a market built on anonymous wallets, and no AI agent needed to prove anything to anyone. A project gets known for the moment it arrives in, and that reputation set early and stuck. Concordium kept building through more than one full market downturn anyway, shipping zero-knowledge identity, euro-pegged fees and two-to-four-second finality while louder projects came and went. That is why the distance between what it is known for and what it runs today is wider than for most of the projects that were louder in 2021.
What changed is the other side of the market. Agents now buy, book and pay, and every one of those actions raises the question Concordium was built to answer: who stands behind this?
The agentic economy did not change the roadmap. It proved it.
Where the CCD Sits
Every identity check, access grant and payment on Concordium is a transaction on the chain, and every transaction is paid for in CCD, the network's native token. Fees are low and stable, pegged to the euro at around a cent, so a business can price an agent's work in advance. Validators and delegators stake CCD to secure the network, and a business can sponsor transactions so its customers and agents never need to hold the token themselves.
The Part Everything Else Rests On
All of this depends on one thing being possible: that a person can prove something true about themselves without handing over the thing that proves it. Age without a passport. Residency without an address. Authority without a signature on file.
Christensen drew that line early, in a 2022 interview: “Privacy is completely different to anonymity. Privacy, I think, is nearly a human right for interactions. Anonymity is not.”
In the next part, the wine order completes in seconds and the retailer never sees a passport.
Frequently Asked Questions
Is Concordium a Layer-1?
Yes, in the sense that matters here: the blockchain underneath it is a public, permissionless Layer-1. Concordium is an AI infrastructure for the agentic economy, and running its own base layer is what allows identity to be enforced when an account is created rather than checked later by an application.
What does Concordium do for AI agents?
It gives a counterparty a way to find out who is accountable for an agent before transacting with it. Every account on the network, including an agent’s, traces back to a verified real-world identity, and an agent can prove an attribute about the person it acts for without exposing their details.
Can an agent running on another chain use this?
Yes, and it can do both at once. An agent keeps running on Ethereum or Solana and registers with Concordium’s agent registry as well, which links it back to a verified owner. Nothing has to move.
Disclaimer: Concordium Explained is published for information only. It is not investment advice or an offer to buy or sell any cryptoasset. Cryptoasset prices are volatile and you could lose all the money you put in.