🧵6/7
Dear Commissioners,
After four hours arguing about how we hold artificial intelligence accountable when these systems start moving real value, writing contracts, executing trades, and settling obligations at machine speed — I want to put one concrete piece of infrastructure on the table: Concordium.
Concordium is a public, permissionless Layer-1 blockchain designed from the ground up around a single non-negotiable principle: every account — whether it belongs to a natural person, a legal entity, or an autonomous AI agent — is anchored to a verified real-world identity at the protocol level. Not as an optional add-on. Not as a compliance middleware someone might bolt on later. It is the condition of entry.
How the identity layer works
A user (or the human principal behind an agent) undergoes a one-time identity verification with an authorized Identity Provider. The resulting identity credentials live primarily off-chain. What the chain itself holds is an encrypted cryptographic link. Day-to-day activity uses zero-knowledge proofs: the account can prove attributes — age, jurisdiction, accreditation, ownership of the agent — without revealing the underlying personal data. Transactions remain private by default.
Identity disclosure is possible, but only through a regulated multi-party legal process involving court orders and independent Privacy Guardians. No single party, including the Concordium Foundation itself, can unilaterally unmask a user. That design deliberately balances the privacy expectations of legitimate market participants with the accountability regulators require when misconduct occurs.
Why this matters for AI accountability
The hardest problem we have spent the afternoon debating is this: when an autonomous agent initiates a transaction, places an order, or settles a payment, who is legally and financially responsible if something goes wrong?
On most public chains today the answer is effectively “no one identifiable.” Concordium answers it at the protocol layer. AI agents are registered through an Agent Registry, minted as non-fungible tokens under standards compatible with emerging agent identity frameworks, and explicitly linked to a verified human or corporate principal. The agent inherits the accountability of its owner. Counterparties can verify that linkage through zero-knowledge proofs without learning the private details. Funds can remain in human custody while the agent is authorized to act within defined bounds. Protocol-level locks and sponsored transactions further constrain risk.
In short, the system is built so that “the AI did it” is never a complete answer. There is always a real-world identity standing behind the agent, recoverable through lawful process.
Operational characteristics relevant to this Commission
• Consensus is ConcordiumBFT, a proof-of-stake Byzantine fault-tolerant protocol delivering deterministic finality in roughly two to four seconds. Once a transaction is finalized it cannot be reversed. That property is material for any market infrastructure that must provide certainty of settlement.
• Transaction fees are designed to be low and stable, denominated with reference to euro terms, reducing the cost unpredictability that has hindered institutional use of other public chains.
• Smart contracts execute in WebAssembly and can be written in Rust, giving developers a relatively mature and auditable environment.
• The same identity primitives support geofencing, jurisdictional controls, and selective attribute disclosure — tools that map directly onto AML, sanctions, and market-integrity requirements.
Concordium is not claiming to solve every regulatory question surrounding AI. It does, however, remove one of the most dangerous structural failures in current public blockchain design: the ability to deploy powerful autonomous economic agents with no protocol-level tether to an identifiable, accountable human or entity.