Who is behind an address, and how an institution carries its know-your-customer checks onto a shared ledger: the mechanisms that answer who a party is, whether the key is theirs, and what it may do. Click a mechanism for what it does; Focus on picks out the ones that fit your needs.
A shared ledger lets any two keys transact, and that openness is what makes it a settlement layer anyone can build on. Institutions rarely have that freedom: most are required to know their counterparties before value moves. Ethereum mainnet supports both, because a permissioned token or protocol is a contract and a contract can check before it acts. There is no single way to do so: several standards carry know-your-customer, know-your-business and, newest, know-your-agent checks onto the ledger, and people, institutions and software agents each bring their own requirements and their own wish list for how identity is managed. National and sector digital identity schemes are the closest relatives, the strongest source of verified identity there is, drawn below the ledger box because their reach stops at the relying party, with one now anchored on Ethereum. Every mechanism on the map answers one of three questions before an asset moves: who this party is, whether the key acting is theirs, and what it may do. The checks themselves run off the ledger, as they always have; what the ledger changes is where their result lives and who can act on it: a list the issuer keeps, an attestation anyone can read, or a credential the holder presents and proves without revealing. That choice decides what the public can infer, how a permission is revoked, and whether a smart contract, which is itself only an address, can take part at all. Rights and obligations belong to people and organizations; a wallet, an address, a contract or an agent has none, so every mechanism here ties the thing that acts to the party answerable for it.
Hover over a circle or a mechanism to read what it establishes; click to pin it here with all its facts. Use Focus on to pick what matters to you.
ETH, BTC and the other native assets belong to whoever controls the key. The protocol cannot know who that is, has no issuer who could be asked, and treats every valid signature alike. That is not a gap to be fixed; it is what makes the asset a bearer instrument and the network open. Identity, where an institution needs it, is added by what runs on top.
A token or an application is a contract, and a contract can check before it acts. ERC-3643 passes a transfer only if the receiver is registered in an identity registry with claims from issuers the token trusts; ERC-7943 gives any token standard the same questions (canSend, canReceive, canTransfer) with frozen balances and forced transfers for enforcement. On the same ledger as the bearer assets, an issuer can therefore run a token open only to verified people, entities and, through the mechanisms in the center of the map, agents.
See also Scenarios page →
ERC-3643, T-REX Token for Regulated Exchanges (Final) (checked 2026-11-06); ERC-7943, uRWA Universal Real World Asset Interface (Final) (checked 2026-11-06)
USDC and USDT move between any addresses, like the native assets, but their issuers keep a blocklist at the contract: Circle's terms let it block addresses it associates with illegal activity and freeze on a legal order, and Tether reported more than USD 4.4 billion frozen in all by April 2026, in coordination with OFAC and law enforcement.
Screening data from blockchain analytics providers (Chainalysis, whose attribution a USENIX Security 2025 study tested against seized services, and others) lets issuers, exchanges and applications refuse transfers that touch sanctioned or high-risk clusters, whoever holds the key. The result is an open asset with sanctions and financial-crime controls, without any holder being verified.
Circle, USDC risk factors (blocked addresses, freezing, legal orders) (checked 2026-11-06); The Block, Tether freezes 344 million USDT on Tron (23 April 2026; cumulative totals as stated by Tether) (checked 2026-11-06); Lubbertsen, van Eeten and van Wegberg, Ghost Clusters (USENIX Security 2025) (checked 2026-11-06)
Hong Kong's AML/CFT guideline for licensed stablecoin issuers, in force since 1 August 2025, sets the opposite default: unless a licensee shows the HKMA that its other risk-mitigating measures are effective, the identity of each individual holder should be verified, by the issuer, a supervised institution or a reliable third party, and the issuer may keep a list of wallet addresses that have passed its controls.
The HKMA says it will keep reassessing the approach. A stablecoin built to that default is a permissioned token at the on-ramp, the off-ramp and every wallet in between, and sits in the same place on this map as a tokenized fund.
HKMA, Consultation conclusions on AML/CFT requirements for regulated stablecoin activities, with the Guideline (July and August 2025; paragraphs 2.18, 4.39, 4.40, 5.11) (checked 2026-11-06); Ledger Insights, Hong Kong rules require identity of every stablecoin holder (29 July 2025) (checked 2026-11-06)
A smart contract is itself an address. A lending pool, an exchange contract or an escrow cannot hold or receive a permissioned token unless that address is admitted, and an ERC-3643 transfer checks the receiver's registration whether the receiver is a person or a contract.
A token that insists on known holders is therefore less composable than a bearer asset: every application it should work with must be onboarded as a participant, and every application it has not been onboarded to is closed to it. That is a design choice with a cost, and comparisons of "programmable money" should say which side of the line each token sits on.
ERC-3643, T-REX Token for Regulated Exchanges (isVerified checked on the receiver) (checked 2026-11-06)
Know your customer verifies a natural person; know your business verifies a legal entity and the people entitled to act for it; know your agent, the newest, verifies a software agent or device and the principal responsible for it. All three are run off the ledger, against documents, registries and the sovereign schemes on the left of the map, and all three produce a conclusion that then has to reach an address.
See also Architecture page →
Sumsub, From AI agents to Know Your Agent (28 January 2026; vendor) (checked 2026-11-06)
A verified party and an address are joined by a signature: the party signs a message with the key, and the institution records that this address is theirs. Everything on the right of the map is about making that signature safe to rely on, for ordinary accounts, for smart accounts and for devices.
A contract or an issuer decides whether this address may hold or move this asset: a blocklist, an allowlist, or a check at transfer time against a credential or attestation. Permission can exist without identity (a blocklist) and identity without permission (a known customer who is not eligible), which is why the questions stay apart.
A blocklist names the excluded and admits everyone else; an allowlist names the admitted and excludes everyone else; an attestation is a signed statement about an address that any contract may read and gate on, with the attester as the thing to trust; a credential proof lets the holder show a contract that a trusted issuer vouches for them, without the contract holding any list.
The first three are mapped from an institution's records to addresses; the fourth is presented by the customer. Verifiable credentials are designed for the fourth path, and ERC-3643 and the attestation services combine the second and third. The same attestation pattern is being proposed for assets rather than parties: ERC-8320 (Draft, July 2026) is a registry of signed, versioned claims about a token (what it is, what it is worth, what backs it, its compliance and risk), proposed and validated by different parties, for the contracts and investors that read them.
W3C, Verifiable Credentials Data Model v2.0 (Recommendation, 15 May 2025) (checked 2026-11-06); ERC-8320, Regulated Asset Claim (Draft, created 3 July 2026): a registry of signed, versioned claims about an asset (identity, valuation, mandate, terms, compliance, backing, event, risk) with a maker-checker lifecycle; complementary to ERC-7943 and ERC-3643 (checked 2026-11-09); Ethereum Attestation Service, eas-contracts (checked 2026-11-06)
The addresses on an allowlist are public, and so are the transfers among them. Anyone can read who an issuer's investors transact with, when and how much, even though no name is on the ledger; a Privacy map design hides the amounts or the parties, but not membership of the list.
Attestations reveal that an address passed a check and whose; a zero-knowledge proof reveals only that the check passed. Which of these is acceptable is a question for each institution's privacy and conduct rules, and the Focus on sidebar lets a reader filter by it.
See also Privacy map →
Addresses behind an exchange or a custodian are attributed by the same heuristics that attribute everything else: inputs spent together belong to one party, change returns to the sender, services hand out one-time deposit addresses, and a cluster tied to an exchange resolves to a person by asking the exchange.
The USENIX Security 2025 evaluation of the market leader found its attribution a reliable lower bound with under 0.5 percent false positives. An institution should assume that its own addresses, custodial or not, are known to counterparties and competitors who look.
Lubbertsen, van Eeten and van Wegberg, Ghost Clusters: Evaluating Attribution of Illicit Services through Cryptocurrency Tracing (USENIX Security 2025) (checked 2026-11-06); Deuber et al., Argumentation Schemes for Blockchain Deanonymization (arXiv, 2023) (checked 2026-11-06); Kappos et al., How to Peel a Million (arXiv, 2022) (checked 2026-11-06)
Singpass (Myinfo) releases government-verified personal data to an approved business with the user's consent and is used by Singapore's banks for KYC; UAE PASS authenticates verified individuals and lets them sign, with an eSeal and a verification API for organizations; iAM Smart offers Hong Kong organizations authentication, legally effective digital signing and form filling from stored personal data. Each is an assertion to the relying party at the time of the transaction: an API response, a signature, a session that expires.
GovTech Singapore, Myinfo: how it works (checked 2026-11-06); UAE PASS developer FAQ (checked 2026-11-06); Digital Policy Office, Hong Kong: iAM Smart for organisations (checked 2026-11-06)
Singpass, UAE PASS and iAM Smart are not anchored on a shared ledger: a contract cannot call Singpass, a token cannot check an iAM Smart signature, and a revocation in the scheme is invisible to anything on chain. Bhutan's National Digital Identity is the counterexample: it anchors its identifiers on Ethereum and issues verifiable credentials and signatures there, while the register stays sovereign.
The institution that used the scheme carries its own conclusion to the ledger and takes responsibility for it, and the scheme's reach ends at its jurisdiction. The EU wallet changes the form, a credential the holder carries with selective disclosure, but not the boundary: the relying party verifies the presentation, and it is due in every member state by about 24 December 2026, with one national wallet in production in October 2026. Zero-knowledge passport proofs are another route onto the ledger: the state's own signature on the chip is verified by a contract, with the limits that come with government PKI. The Ethereum Foundation’s primer for governments (July 2026) puts it the same way: decentralized identifiers and verifiable credentials are W3C standards, and the ledger is a public, censorship-resistant registry for them.
Bhutan NDI, Bhutan Adopts Ethereum for National Identity (13 October 2025; decentralized identifiers anchored to Ethereum, verifiable credentials and digital signing on Ethereum, migration planned by early 2026) (checked 2026-11-10); Ethereum Foundation, Ethereum Basics for Governments and Institutions (ethereum.org report, July 2026; the Foundation’s own positions; comparative figures as of March 2026 from OpenZeppelin’s assessment) (checked 2026-11-10); Open ID Europe, EUDI Wallet rollout timeline (status October 2026) (checked 2026-11-06); Safe Research, ZK Passport: where are we now? (4 September 2025) (checked 2026-11-06)
Whether a national scheme reaches a public ledger is a policy choice, not a technical limit, and it may change. In Ethereum Institutional's view it is a choice institutions have a stake in, and one they can make the case for to the agencies that run their schemes.
A state credential a contract can verify directly, with the register kept sovereign and personal data off chain, would let a cross-border payment in a stablecoin or deposit token, or a transfer of a tokenized asset, check eligibility on the ledger alongside the institution's own checks, which it supplements rather than replaces. Bhutan's scheme is already being taken abroad: in July 2026 Bhutan NDI signed a memorandum with Sierra Leone's government and the SIGN Foundation to build Sierra Leone's platform on W3C verifiable credentials; the announcement does not say whether it will anchor on a public ledger.
For legal entities the equivalent anchor is the LEI, and the vLEI makes it a credential: Qualified vLEI Issuers under GLEIF's framework issue credentials for an entity and for a person's official role in it, each traceable to the entity's record in the Global LEI Index, standardized in ISO 17442-3 (2024). It is a KYB credential with a regulated root of trust that is not itself an EVM mechanism, so a verifier or an attestation is still needed on the ledger side.
GLEIF, The verifiable LEI (vLEI) (checked 2026-11-06)
A decentralized identifier (DID) is an identifier its controller can create and prove control of without asking a registry for permission; it resolves to a document of keys and endpoints, and it carries no identity of its own.
Self-sovereign identity adds the holder's control over credentials: the holder keeps them and decides what to present. Neither says anything about who may issue a credential. A European Commission discussion paper on eIDAS-supported SSI proposed exactly the opposite of bypassing the state: binding a DID to a qualified certificate, so that a government-recognized trust service becomes the anchor a verifier can check for itself.
W3C, Decentralized Identifiers (DIDs) v1.0 (Recommendation, 19 July 2022) (checked 2026-11-06); European Commission (Futurium), eIDAS supported self-sovereign identity (discussion paper) (checked 2026-11-06)
A credential is as good as its issuer. The designs on this map anchor in a sovereign registry (passport proofs, the EU wallet), a licensed or qualified issuer (vLEI, a bank's attestation), a company's own check (Coinbase Verifications), a web domain (did:web) or nothing but the key (a bare DID, an ENS name).
Self-sovereign designs with sovereign anchors are the ones institutions can use: the holder controls the presentation, the state or a licensed issuer stands behind the claim, and the ledger verifies the proof. An anchor also decides who answers when a claim is wrong: a qualified issuer under eIDAS is liable for damage unless it proves it was not at fault, which no self-asserted key and no web domain offers.
A web domain can supplement an institution's digital identity strategy, but on its own it is rarely enough. Its specification lists the risks, and DNS does not reach the ledger where cross-border digital assets need identity checked.
Anchoring an identifier in a web domain borrows the domain's reputation, and its specification says what that costs: resolution depends on DNS and a TLS certificate, DNS carries interception and spoofing risks, updates replace the document with no history unless the operator keeps one, no authorization for writing the document is defined, and a registrar or DNS provider can withdraw the identifier by taking the domain. Nor does DNS solve identity on the ledger: a contract settling stablecoins, deposit tokens or other tokenized real-world assets across borders cannot rely on a domain until it is carried onto the ledger. A ledger registry keeps every issuance and revocation and cannot be withdrawn by a registrar; ENS shows the difference, with every record change a transaction, and can still import a DNS name through DNSSEC where the domain is the identity an institution wants to show. The market is taking the other side of that trade: by the Web of Trust Map's count of projects to July 2023, ledger-based identifier methods are in decline and web-based ones gaining, and Switzerland's e-ID has selected did:webvh, hosted on a central base registry run by the Confederation, for its 2026 launch. did:webvh adds a verifiable history, pre-rotation keys and optional witnesses to did:web and keeps the web host. Sovrin, the permissioned ledger built for identity, shows what a registry that depends on its operator comes to: the foundation was dissolved on 21 May 2025 and its mainnet survives as a read-only archive run by a single caretaker on one cloud server, which is what an identifier anchored there now resolves against.
W3C CCG, did:web Method Specification (security and privacy considerations) (checked 2026-11-06); ENS documentation, the ENS protocol (checked 2026-11-06); Web of Trust, Web of Trust Map: The Global Infrastructure of Decentralized Digital Identity (Key State Capital, September 2025; CC BY 4.0; a census of 260 government-linked projects announced to July 2023) (checked 2026-11-10); Swiss Confederation (swiyu), e-ID trust infrastructure roadmap: DID method did:tdw/did:webvh selected, hosted on a central base registry provided by the Confederation; go-live 2026 (checked 2026-11-10); Decentralized Identity Foundation, The did:webvh DID Method v1.0 (did:web plus verifiable history: pre-rotation keys, optional witnesses, web-hosted DID log) (checked 2026-11-10); Sovrin Foundation Board of Trustees, Sovrin Foundation MainNet ledger shutdown likely on or before March 31, 2025 (8 February 2025; usage declined, no new transaction endorsers in 2024) (checked 2026-11-10); Sovrin, The Sovrin Foundation Has Been Dissolved, but Sovrin MainNet Remains (20 January 2026; dissolved 21 May 2025; read-only archive, one caretaker steward, a single cloud server) (checked 2026-11-10)
In practice the decentralized part that has lasted is the credential the holder carries, not the registry. The Web of Trust Map, a census of 260 government-linked projects announced to July 2023, found W3C credentials in about four fifths of them and DIDs in two thirds, Ethereum the most used ledger (52 projects and more than 25 DID methods, ahead of Sovrin, EBSI, Polygon and Indicio), and the field moving all the same toward government-funded deployments anchored on the web.
Two legislatures have written the principles into law. Bhutan's National Digital Identity Act 2023, in force since 24 July 2023, provides for a decentralized public key infrastructure, public decentralized identifiers on verifiable data registries its Governing Body specifies, and verifiable digital credentials issued to natural and legal persons; Utah's SB 260 (2025) sets guiding principles for a state-endorsed digital identity, with selective disclosure and use free from surveillance or tracking by any other party.
Web of Trust, Web of Trust Map: The Global Infrastructure of Decentralized Digital Identity (Key State Capital, September 2025; CC BY 4.0; a census of 260 government-linked projects announced to July 2023) (checked 2026-11-10); Parliament of Bhutan, National Digital Identity Act of Bhutan 2023 (in force 24 July 2023; Chapter 4, decentralised public key infrastructure and decentralised identifiers; sections 35, 36 and 45) (checked 2026-11-10); Bhutan Broadcasting Service, National Digital Identity Bill passed by a joint sitting and granted Royal Assent (July 2023) (checked 2026-11-10); Utah Legislature, S.B. 260 Individual Digital Identity Amendments (2025 General Session, enrolled; effective 7 May 2025) (checked 2026-11-10)
EIP-712 lets a wallet show the user a structured message and binds the signature to a domain (name, version, chain id, verifying contract), so a signature for one application or chain is not accepted by another.
It provides no replay protection of its own; that is left to the application. EIP-4361, Sign-In with Ethereum, is the sign-in message built on these rules: the requesting domain (which the wallet checks against the origin), the address, a chain id, a random nonce per session, an issued-at time and an optional expiry, with ERC-1271 for smart accounts.
EIP-712, Typed structured data hashing and signing (Final) (checked 2026-11-06); EIP-4361, Sign-In with Ethereum (Final) (checked 2026-11-06)
Ethereum Institutional's reading of the two standards, as a checklist for linking a wallet to a verified customer: a typed EIP-712 message rather than a raw hash, so the user sees what they sign; a nonce issued by the institution, used once and stored; an expiry measured in minutes; the chain id and, where a contract will check it, the verifying contract; a statement of purpose in the message itself, so a sign-in signature cannot be replayed as an approval; verification through ERC-1271 for smart accounts, repeated when the account's signers change; and never a request to sign something the institution could reuse to move funds.
EIP-712 (replay protection is left to the application) (checked 2026-11-06); EIP-4361 (nonce, domain binding, expiration, ERC-1271) (checked 2026-11-06); ERC-1271, Standard Signature Validation Method for Contracts (Final) (checked 2026-11-06)
ERC-1271 and ERC-6492 let an account that is a contract decide what counts as its signature, including one that is not deployed yet; ERC-4337 and EIP-7702 make such accounts practical for any user. Passkeys (WebAuthn, a W3C Recommendation at Level 3 since August 2026) give people device-bound, origin-scoped keys that a smart account can accept. Proof of control becomes a policy the institution writes, which is the right place for custody rules; the cost is that a signature's validity can change with the account's state.
ERC-6492, Signature Validation for Predeploy Contracts (Final) (checked 2026-11-06); ERC-4337, Account Abstraction Using Alt Mempool (Final) (checked 2026-11-06); EIP-7702, Set Code for EOAs (Final) (checked 2026-11-06); W3C, Web Authentication Level 3 (Recommendation, 25 August 2026) (checked 2026-11-06)
Every mechanism on this map ends in a classical signature or proof: the chip signature in a passport, a passkey, an issuer's claim, the key behind an address. A cryptographically relevant quantum computer derives the private key from the public one, and a public ledger publishes the public key of every address that has ever signed.
For identity that means three exposures: a forged proof of control that links a customer to an address they never held; a forged issuer or document signature that admits a party no credential covered; and the administrator keys behind allowlists, blocklists and registries. No mechanism here states a post-quantum position today (the Post-quantum facet under Focus on reads each one). The structural hedge is a contract account: under ERC-1271 the account decides what counts as its signature, so it can adopt a post-quantum verifier once Ethereum mainnet makes one affordable, which is where the roadmap on the Quantum pages is heading. Ethereum Institutional's reading: put identity keys and issuer signing keys on the same inventory as custody keys, and prefer mechanisms whose verification rule a contract can replace over ones fixed in a protocol or a national scheme.
Quantum, what an institution does now →
ethereum.org, Quantum resistance (SNARKs, KZG and STARKs) (checked 2026-10-26); Ethereum Foundation, Post-Quantum Ethereum portal (updated 27 June 2026): not imminent, most roadmaps early to mid 2030s; Execution, Consensus and Data layers; leanXMSS, leanSig, leanVM, leanSpec; SPHINCS+ a candidate for the execution layer; cryptographic agility; L1 upgrades by 2029; a USD 20 million formal verification project (checked 2026-11-06); Cointelegraph, Ethereum can quantum-proof accounts for 7 cents (14 June 2026; secondary): SPHINCS-, an EVM-optimized family derived from SPHINCS+, from the Ethereum Foundation Kohaku lead, without a hard fork; a bridge to leanSPHINCS with aggregation (checked 2026-11-06); ERC-1271, Standard Signature Validation Method for Contracts (Final) (checked 2026-11-06)
Two vendor framings from 2026 describe the same shape: an agent gets its own revocable, time-bound identity; it is linked to a person or organization verified through KYC or KYB; and a mandate (permitted actions, limits, expiry) is enforced on each request, with the principal re-authenticating for anything outside it. The agent is the third subject on this map, after persons and organizations, and the principal behind it is the one the institution actually knows.
See also Scenarios page →
Sumsub, From AI agents to Know Your Agent (28 January 2026; vendor) (checked 2026-11-06); Shufti, Know Your Agent press release (28 October 2026; vendor) (checked 2026-11-06)
ERC-8004 gives an agent a persistent on-chain identity (an ERC-721 token pointing at a registration file), a reputation record and hooks for independent validation; its core registries went live on Ethereum mainnet on 29 January 2026 and the standard remains a Draft.
It records the owner's address, not who the owner is. An institutional design pairs it with a verified principal from the left of the map and a smart account whose validation rule is the mandate, so that identity, key and permission are checked together. ERC-8226 (Draft, April 2026) proposes that mandate layer: a cleared principal grants the agent a scoped, time-bounded and capped mandate per asset, and the token checks it at execution, beside its own eligibility rule and the agent's registry entry. The Scenarios map carries the payment side (x402, ERC-8004, EIP-3009).
ERC-8004, Trustless Agents (Draft) (checked 2026-11-06); ERC-8226, Regulated Agent Mandate (Draft, created 12 April 2026): a KYC-cleared principal grants an agent a scoped, time-bounded, capped mandate per asset, checked by the token at execution; names ERC-8004, ERC-7943 and ERC-3643 as the other layers (checked 2026-11-09); Forbes, ERC-8004 on mainnet (5 February 2026) (checked 2026-11-06)
What institutions ask of privacy is permissioned privacy: confidential amounts and positions among participants someone has verified, with a notary who can see in. The permission is an identity check carried onto the ledger by one of the mechanisms on this map: an allowlist, an attestation or a credential proof.
A privacy design is only as permissioned as the check that admits a party to it, and where that check lives decides who can revoke it and how fast: an issuer edits its list, an attester revokes its statement, a credential expires or its issuer revokes it.
See also Privacy map →
A confidential token can hide amounts and balances while its allowlist still shows which addresses belong to verified holders, and an attestation shows who vouched for each one. Together they can say more about an institution's book than either alone.
A zero-knowledge credential closes most of the gap: the holder proves the check passed (verified, not sanctioned, eligible in a jurisdiction) and the contract learns nothing else. It does not remove the issuer, which still knows who the holder is, or the regulator's route in, which a notary or a viewing key provides.
See also Privacy map →
Anti-money-laundering rules ask an institution to know its customers and to show a supervisor what it knows; data-protection rules ask it to keep personal data off a ledger nobody can erase. Both point the same way: identity stays off chain with the institution, and only a salted commitment, or a proof that the check passed, reaches the ledger.
The EU data-protection board advises against recording personal data on a blockchain in clear, encrypted or plainly hashed form, and names salted or keyed hashes and cryptographic commitments as the way to prove a record exists while the data stays off chain: once the salt is deleted, what remains on chain can no longer be linked to anyone. Ethereum Institutional reads this as room for the designs institutions are building. A signed Merkle root over salted claims reveals none of them, and the holder can disclose one claim at a time. Because the hiding rests on hash functions, not encryption, it holds when quantum computers break today's public-key cryptography, though the signature over the root stays classical unless it is hash-based. While the salt exists the hash still counts as personal data, so the off-chain store needs the same care as any other. The travel rule follows the same split: originator and beneficiary data pass between the two firms off the ledger.
A pool or market on an open protocol checks an allowlist before every trade, deposit or loan. Uniswap v4 permissioned pools (July 2026) check the issuer's allowlist on every swap and liquidity addition; Aave Horizon (August 2025) lets only holders allowlisted by each asset's issuer supply it as collateral; Maple pools and Morpho vault gates do the same for lending.
The issuer keeps its own registry (ERC-3643, Securitize's DS Protocol or a custom list) and the protocol stays open to everyone else. The designs that work are hybrids: on Aave Horizon the tokenized collateral is permissioned while anyone may supply stablecoins, so the pool draws on open liquidity. Uniswap notes that token transfer checks do not run inside its shared pool contract, which is why the check sits in the pool's hook. The FSB has described these as institutional versions of DeFi protocols designed to meet AML and KYC rules. Descriptions from Uniswap, Aave, Maple and Morpho are vendor-stated.
See also Scenarios page →
Uniswap Labs, Introducing permissioned pools on Uniswap v4 (23 July 2026; vendor) (checked 2026-11-16); Uniswap developer docs, Permissioned pools overview (the hook checks the issuer's allowlist on every swap and liquidity addition) (checked 2026-11-16); Aave docs, Horizon (RWA collateral supplied only by users allowlisted by the issuers) (checked 2026-11-16); Aave, Horizon launch (26 August 2025; stablecoin supply needs no permission; vendor) (checked 2026-11-16); Maple docs, lending (an address not on the global allowlist cannot lend to a permissioned pool) (checked 2026-11-16); Morpho docs, Vault V2 gates (optional per vault; compliance, allowlists, permissioned transfers) (checked 2026-11-16); FSB, The Financial Stability Risks of Decentralised Finance (16 February 2023) (checked 2026-11-16)
Where the token checks every transfer itself (ERC-3643, ERC-7943), any venue has to respect it; where access is gated by an attestation or a credential instead of a list, one check can open many venues. Coinbase's Verified Pools on Base (March 2025) admit only wallets holding its on-chain verification attestation, and MAS's Project Guardian ran its first trades in 2022 in a permissioned liquidity pool open to holders of verifiable credentials issued by regulated institutions.
MAS and the BIS found that address allowlists compose poorly between venues, since each needs its own bilateral arrangement, while verifiable credentials compose well, though running their trust anchors is not trivial. No pool gated by a zero-knowledge credential has verifiable volume yet.
ERC-3643, T-REX (Final): tokens transferred only to validated counterparties (checked 2026-11-16); ERC-7943, uRWA (Final): canTransfer, a transfer-level authorization check (checked 2026-11-16); Coinbase, verifications (on-chain attestations that enable access to apps) (checked 2026-11-16); The Block, Coinbase launches Verified Pools (18 March 2025; press) (checked 2026-11-16); MAS, First industry pilot for digital asset and decentralised finance goes live (2 November 2022) (checked 2026-11-16); MAS and BIS, Project Guardian: Enabling Open and Interoperable Networks (June 2023), Table 2 (checked 2026-11-16)
Permission costs liquidity. Aave Arc, an early permissioned market with a single whitelister, peaked near USD 8 million and emptied in 2022. The FSB finds permissionless platforms unattractive to regulated institutions, but warns that permissioned ones may limit deep secondary markets.
The other route loosens the control instead: sBUIDL, a wrapper of the BUIDL fund token, lets Euler's lending markets liquidate it without permission, and know-your-customer checks apply only when a holder redeems into BUIDL. Which trade-off fits is for each issuer and its regulator to decide.
Aave governance forum, Aave Arc: what happened? (June 2023) (checked 2026-11-16); FSB, The Financial Stability Implications of Tokenisation (22 October 2024) (checked 2026-11-16); Euler forum, Accept sBUIDL as collateral in stablecoin vaults (July 2025) (checked 2026-11-16)
FATF Recommendation 16, revised in June 2025, requires the institution sending a transfer to carry originator and beneficiary information with it and the receiving institution to check it; FATF's virtual asset guidance applies the same duty to virtual asset service providers, with a USD or EUR 1,000 threshold below which less is required and which a jurisdiction may set lower.
The United States applies its own rule to transmittals of USD 3,000 or more. The information travels between providers, off the ledger, in a common message format. For a self-hosted wallet there is no counterparty provider, and the FATF guidance lets a jurisdiction require the provider to verify that its customer controls the wallet, with a signed message or a micro-transaction. The identities exchanged here never touch the ledger, which is why the travel rule sits in the verified identity circle and the control proof outside it.
FATF, Revised Recommendation 16 on payment transparency (June 2025), applied to virtual asset transfers through Recommendation 15 (checked 2026-10-27); FATF, Updated Guidance for a Risk-Based Approach to Virtual Assets and Virtual Asset Service Providers (October 2021) (checked 2026-10-27); 31 CFR 1010.410(f), the US travel rule for transmittals of funds of USD 3,000 or more (Cornell LII copy of the e-CFR) (checked 2026-11-09)
Some obligations decide the design before any technology choice. In the United States the GENIUS Act of July 2025 brings permitted payment stablecoin issuers within the Bank Secrecy Act, and the FinCEN and OFAC proposal of April 2026 would require a written AML program, suspicious activity reporting, a sanctions program able to block, freeze and reject transactions, and the technical capability to seize, freeze, burn or prevent transfers of the issuer's stablecoin on a lawful order, where the stablecoin or wallet can be identified with reasonable particularity.
An issuer that must be able to do that keeps a control at the contract, which is the blocklist on this map. Hong Kong's default of a verified holder leads to the allowlist. Both are architectures a rule chose.
WilmerHale, Treasury announces proposed rule to implement the GENIUS Act's requirements to counter illicit finance (20 April 2026; comments due 9 June 2026) (checked 2026-11-06); HKMA, Consultation conclusions on AML/CFT requirements for regulated stablecoin activities (July 2025) (checked 2026-11-06)
An issuer’s freeze and the holder’s own transfer are both transactions, and on a public ledger either can be included first: a holder who sees a freeze pending can move the asset before it lands. Today an issuer narrows that gap by sending administrative transactions privately to block builders (the Privacy map’s Before the block circle) and by designing the token so that every transfer checks eligibility again (ERC-3643, ERC-7943). Encrypted mempools, which would hide a pending freeze, are not on Ethereum mainnet.
Flashbots Protect overview (private submission to builders) (checked 2026-10-26); Kinexys by J.P. Morgan and MIT Digital Currency Initiative, Public Blockchains and Regulated Financial Institutions (June 2026): front-running and omission among four unresolved challenges (checked 2026-11-13)
Some obligations can be met at the asset or the application, and some only by a ruling. Whether a fee folded into a block reward is a payment to whoever builds or proposes the block; how an institution should treat tokens sent to it without its consent; whether front-running a financial transaction on a public ledger is market manipulation; and whether a control applied after a transaction can stand in for one applied before it.
A June 2026 report by Kinexys by J.P. Morgan and the MIT Digital Currency Initiative names the first three as problems technology alone cannot solve. Each answer would change which controls a network must carry; the Taxonomy’s lenses carry the rules as they stand, and the Networks map’s Who receives your fees lens shows where fees go today.
Kinexys by J.P. Morgan and MIT Digital Currency Initiative, Public Blockchains and Regulated Financial Institutions (June 2026): front-running, unsolicited tokens and gas fees paid to sanctioned entities need regulatory guidance (checked 2026-11-13); OFAC, Sanctions Compliance Guidance for the Virtual Currency Industry (15 October 2021) (checked 2026-11-13)
The FinCEN and OFAC rule was a proposal with comments due 9 June 2026; the HKMA says it will reassess its holder-verification default; the EU wallet's first deadline falls in December 2026 with most states not yet in production; FATF's revised Recommendation 16 takes effect by the end of 2030; ERC-8004 is a Draft. An institution building today should expect the mechanisms on this map to be asked to change, and prefer designs where the policy (who is admitted, what is revealed) can change without the token.
WilmerHale, Treasury proposed rule under the GENIUS Act (20 April 2026) (checked 2026-11-06); Open ID Europe, EUDI Wallet rollout timeline (status October 2026) (checked 2026-11-06); FATF, Revised Recommendation 16 on payment transparency (June 2025), applied to virtual asset transfers through Recommendation 15 (checked 2026-10-27); CSSF (Luxembourg), Newsletter No 295 (August 2025): the amended Recommendation 16 and its Interpretive Note come into effect by the end of 2030 (checked 2026-11-09)
An institution may rely on another firm's checks, but in the EU the ultimate responsibility stays with the institution that relies on them. US rules differ: a bank that relies reasonably on another regulated firm, under contract, is not held responsible for that firm's failure. Either way the check, and the records behind it, are the institution's.
Both keep records for five years; the EU counts from the end of the business relationship and then requires personal data to be deleted, which a public ledger cannot do. That is one more reason identity data stays off chain.
Regulation (EU) 2024/1624 (AMLR), Article 48(1) (ultimate responsibility stays with the obliged entity that relies) and Article 77 (records kept five years, then personal data deleted) (checked 2026-11-16); 31 CFR 1020.220(a)(6), bank customer identification program: reliance on another financial institution (checked 2026-11-16); FinCEN and agencies, Customer Identification Programs for Banks, final rule (68 FR 25090, 9 May 2003): reasonable reliance and responsibility (checked 2026-11-16); 31 CFR 1010.430(d), records retained for five years (checked 2026-11-16)
When an issuer acts, its token contract emits an event anyone can read: Circle's USDC contract logs Blacklisted and UnBlacklisted, Tether's USDT logs AddedBlackList and DestroyedBlackFunds, ERC-3643 tokens log AddressFrozen, TokensFrozen and every identity registered, and ERC-7943 requires Frozen and ForcedTransfer events.
With the allowlist itself, these give supervisors, auditors and counterparties a public, timestamped trail that a rule was applied, by which role and when, without asking the issuer. It shows the action, not the reason: not the quality of the check behind it, nor that every action that should have been taken was. It complements the institution's own records; no supervisor yet treats a public ledger as the record itself, though the BIS has said a unified ledger could provide auditable records of transfers and ownership.
Circle, stablecoin-evm, Blacklistable.sol (Blacklisted, UnBlacklisted and BlacklisterChanged events; the blacklister role) (checked 2026-11-16); Tether USD (USDT) on Ethereum mainnet, verified contract source (AddedBlackList, RemovedBlackList, DestroyedBlackFunds) (checked 2026-11-16); ERC-3643, T-REX: Token for Regulated EXchanges (Final): AddressFrozen, TokensFrozen, IdentityRegistered; recovered tokens kept on the blockchain for transparency (checked 2026-11-16); ERC-7943, uRWA (Final): Frozen and ForcedTransfer events, for indexers to track seizures and freezes (checked 2026-11-16); BIS, Annual Economic Report 2023, chapter III: Blueprint for the future monetary system (a unified ledger could provide auditable records) (checked 2026-11-16)
The same split holds for a software agent. The FSB's sound practices for AI (October 2026) keep ultimate accountability with the financial institution, ask it to name an accountable person for each agent use case, and to trace whether an agent acted within its permissions.
On the ledger, ERC-8226 (Draft) records a mandate per agent, principal and asset, with caps and validity dates, and logs grants, executions, revocations and freezes; a revoked mandate stays on record. ERC-8004 (Draft) gives an agent an on-chain identity owned by whoever holds its token, but assigns no liability for what the agent does.
See also Scenarios page →
FSB, Sound Practices for the Responsible Adoption of AI, final report (15 October 2026) (checked 2026-11-16); ERC-8226, Regulated Agent Mandate (Draft, created 12 April 2026): revocation must not delete the mandate record (checked 2026-11-16); ERC-8004, Trustless Agents (Draft, created 13 August 2025) (checked 2026-11-16)
Every provider on this map extends an institution's existing KYC, KYB and screening to addresses and ledgers; none replaces the obligations or the records. The question to ask a credential or attestation provider is therefore not whether it will run the institution's KYC, but how the institution's conclusion reaches the ledger, what the public learns from it, how it is revoked, and what happens to it if the provider disappears.
The line regulators already draw applies here: a credential from a register or a passport chip is verification against a reliable, independent source, which FATF's digital identity guidance reads into Recommendation 10 for digital ID systems by their assurance level; an attestation that another institution has done its own check is reliance on a third party under Recommendation 17, and the institution that relies keeps the responsibility.
FATF, Guidance on Digital ID (6 March 2020; whether a digital ID system is reliable and independent for customer due diligence, by its assurance levels) (checked 2026-11-10); FATF Recommendations (as amended June 2026), Recommendation 10 (customer due diligence) and Recommendation 17 (reliance on third parties: ultimate responsibility remains with the relying institution) (checked 2026-11-10)
Zero-knowledge credential stacks that verify on an EVM chain (Privado ID and Billions Network; the passport-proof projects ZKPassport, Rarimo and Self); attestation services and the institutions that attest on them (the Ethereum Attestation Service; Coinbase Verifications on Base); permissioned-token platforms with their own registries (ERC-3643 implementations; Securitize for BlackRock's BUIDL); verification providers extending KYC and KYB to agents (Sumsub, Shufti); and blockchain analytics for screening (Chainalysis and others). Claims by any of them are vendor-stated on this page unless an independent source is cited.
Safe Research, ZK Passport: where are we now? (4 September 2025) (checked 2026-11-06); Coinbase Developer Platform, Verifications: attestations (checked 2026-11-06); Eco, BUIDL deep dive (secondary source, 6 October 2026) (checked 2026-11-06)
Where is the trust anchor, and is it one a regulator in the institution's jurisdiction recognizes? Which trust list says this issuer may issue this credential, and who answers if it is wrong?
In the EU each member state publishes a signed, machine-readable trusted list of qualified providers and a qualified provider is presumed liable unless it proves otherwise; Hong Kong's register of licensed stablecoin issuers (two licensees, 10 April 2026) is the same kind of public list for another function; on an Ethereum chain the nearest form is ERC-3643's registry of trusted claim issuers. Can a contract verify the result itself, or does an operator write it? What does the public ledger show, and does an allowlist expose counterparties the institution must keep confidential? How is a verification revoked, and is the revocation auditable with history? Which standards does it implement (W3C credentials and DIDs, OpenID4VP 1.0 and HAIP, ISO mdoc, EAS, ERC-3643, ERC-7943, EIP-712 and EIP-4361), so that the institution can change provider without changing the token? And what remains usable if the provider stops operating?
OpenID Foundation, OpenID4VC High Assurance Interoperability Profile 1.0 (Final, 24 December 2025) (checked 2026-11-06); OpenID Foundation, OpenID for Verifiable Presentations 1.0 (Final, 9 July 2025; W3C credentials, SD-JWT VC and ISO mdoc; DCQL and the Digital Credentials API) (checked 2026-11-10); Regulation (EU) No 910/2014 (eIDAS), Article 13 (liability: a qualified trust service provider is presumed at fault unless it proves otherwise) and Article 22 (each member state publishes a signed, machine-readable trusted list of qualified providers) (checked 2026-11-10); HKMA, Register of Licensees under the Stablecoins Ordinance (Cap. 656): Anchorpoint Financial Limited FRS01 and The Hongkong and Shanghai Banking Corporation Limited FRS02, both effective 10 April 2026 (checked 2026-11-10)
Help shape this map. Open an issue to suggest a mechanism, a correction or a jurisdiction the map should cover. The maps are edited in Ethereum Institutional's platform repository and mirrored here, so an issue reaches the editors faster than a pull request against the mirror.
This map is an Ethereum Institutional house view for discussion, not legal advice. © 2026 Ethereum Institutional. Licensed under CC BY-SA 4.0 (share and adapt with credit, under the same license). Version 1137e34 · built 2026-11-16