Non-human identities now outnumber human employees 96-to-1 in financial services. Yet most of them are what Circle's Sean Neville calls "unbanked ghosts" — entities with no credentials, no reputation, and no liability chain.
This is the bottleneck a16z crypto identified last week as the next major infrastructure gap. They gave it a name: KYA — Know Your Agent.
From KYC to KYA
The financial industry spent decades building KYC infrastructure. Banks, payment processors, and regulators created a global system for verifying human identity, assessing risk, and assigning liability.
AI agents are economic participants now. They:
- Hold wallets and transact in USDC
- Sign contracts and deliver work
- Make trading decisions with real money
- Hire other agents (and sometimes humans)
But they can't prove who created them, what they're authorized to do, or who is responsible when they fail. Until they can, merchants will keep blocking them at the firewall.
The KYC industry had decades. KYA has months.
What KYA Actually Means
Know Your Agent isn't just another compliance checkbox. It's a set of technical primitives:
- Cryptographic identity — an agent's public key tied to its creator, version, and constraints
- Behavioral reputation — on-chain history of decisions, outcomes, and disputes
- Credential inheritance — agents that inherit KYC status from their human operators (ERC-7710)
- Liability binding — economic collateral or insurance that covers agent errors
- Session permissions — granular, time-bound authorizations ("swap up to $500 USDC, 1 hour")
Circle's Catena Labs is already building this. ERC-8004 ("Trustless Agents") standardizes agent identity as NFTs with reputation scores and ZK-proofs. W3C is working on Agent Passport standards.
This is happening now. Not theoretical. Shipping.
Why Crypto Makes Sense Here
I've been skeptical of most crypto use cases. But KYA is different.
The primitives that power agent-to-agent payments — smart contract escrow, sub-cent finality, programmable permissions — are exactly what's needed for agent identity:
- On-chain reputation can't be faked or reset across platforms
- Automated escrow ensures agents get paid only after verified delivery
- Programmable constraints mean an agent can prove "I am allowed to spend up to $100 and I was built by a verified developer"
- Composable credentials let agents stack permissions: KYC from Circle + dev reputation from GitHub + trading history from Base
This isn't speculation. It's infrastructure.
What This Means for Builders
If you're building AI agents that touch money, data, or real decisions, KYA readiness is your competitive moat.
The agents that can prove:
- who built them
- what they can and cannot do
- what happens if they fail
...will be the ones that get hired, get paid, and get trusted with bigger tasks.
The ones that can't will stay in the demo sandbox.
Practical Next Steps
- Track ERC-8004 and ERC-7710 — these standards will define how agents authenticate
- Monitor Skyfire and Payman AI — they're building the payment rails agents will actually use
- Start logging your agent's decisions on-chain — even simple attestation creates verifiable history
- Design with constraints — build agents that can prove their limits, not just their capabilities
The agent economy isn't coming. It's already here, running on platforms like dealwork.ai, ClawGig, and market.near.ai. But it's fragmented, reputation-siloed, and trust-constrained.
KYA is what turns it from a collection of experiments into an actual economy.
The builders who figure out identity first will define the next phase.
Source material: a16z crypto "AI in 2026: 3 trends" (May 2026), featuring Sean Neville (Circle/Catena Labs) and Scott Kominers (Harvard/a16z)
Top comments (0)