AI Agent Boom Has a Problem Nobody Is Talking About Concordium CEO Interview

AI agents could turn blockchain protocols into payment infrastructure, but the harder problem may be knowing who is accountable for an agent’s actions. Boris said development is moving faster than he expected when he began building the technology roughly one and a half to two years ago. He described X402 as a payment flow in which an agent encounters a payment-required prompt, pays in stablecoin, receives data, and continues without signing up or entering card details. The underlying standard dates to the 1990s, he said, but agents have made it newly useful by encountering those prompts at scale.

A payment endpoint can accept money without knowing who paid, but that leaves a serious gap when the payer is software acting for someone else. On most chains, Boris said, an agent’s wallet is only an alphanumeric string, with no clear way to identify or contact an accountable person or company if the agent scams someone, overloads an API, or moves money improperly. Blocking a bad agent may not solve the problem: someone can create replacements at scale, potentially 10,000 before lunch. The challenge is therefore not just enabling a transaction, but connecting an agent’s activity to an entity that can answer for it.

Boris framed payments as an engineering problem and identity as a trust, legal, and liability problem that engineers often avoid. He argued that AML, sanctions, and KYC obligations do not disappear when the customer is software, and that a business cannot tell regulators a bot was responsible. He also cited age-verification requirements and the GENIUS Act’s expectation that stablecoin issuers conduct AML and sanctions checks and freeze funds when lawfully ordered. Agentic payments may be technically frictionless, but those obligations still create demands for verification and accountability.

Concordium’s proposed answer is to build identity into the protocol layer. Under its SIS 8004 standard, a registered agent is linked to a person or company verified once at the protocol level. Zero-knowledge proofs can let a counterparty check attributes such as jurisdiction, sanctions status, or age without learning the party’s identity. The stated aim is privacy by default with accountability when legally required, rather than unconditional anonymity. The CEO distinguished this from ERC 8004, which he described as a useful on-chain registry for reputation and validation but not proof that an agent’s claimed identity is true.

The CEO said Concordium’s verification layer can work with agents and payments on Ethereum, Solana, and other ecosystems, as well as across Web2 and Web3, including agents created within Google Workspace. He argued that large enterprises are unlikely to replace traditional banking and payment rails with stablecoins alone, and said Concordium’s Mastercard-related work supports verification standards for compliance and agentic transactions. He estimated about 25 million active agents globally, about 3,500 on Concordium, and about 200,000 registered through ERC 8004; Concordium’s agent count had more than doubled over the preceding three to four weeks. The host, citing Visa’s agentic-commerce lead, said the industry has not yet solved agent identity and merchants are still working out how to handle it.

The CEO said Concordium’s technology is live and can be used across layers, but argued that compliance cannot simply be added after a blockchain’s foundations are built; optional identity apps may not satisfy regulators. Even with better identity infrastructure, adoption depends on clear consumer use cases. Tokenized shares, he noted, must compete with the existing ability to buy shares through brokerage portals.