A Rule That Names a Token, Not Only a Company
On 2026-10-05, FinCEN published a proposed rule at 91 FR 63208, Federal Register document number 2026-20371, that would prohibit covered financial institutions from engaging in a transmittal of funds involving A7 Network Sub-Agents. The filing is public in the Federal Register entry for document 2026-20371. The prohibition has a familiar shape. What an engineer should notice sits elsewhere: the rule does not stop at naming a company, it names a token, and in a footnote two smart contract addresses, one on Ethereum and one on Tron.
That difference carries more than the prohibition. A company is an entity a human analyst can describe in prose; a contract address is a string software can query without knowing its owner. Printing both points a federal document at objects on a public ledger. Comments close 2026-11-04, which is 30 days after publication.
What the Rule Prohibits, and the Shape of the Obligation
Proposed 31 CFR 1010.668(b)(1) bars a covered financial institution from engaging in a transmittal of funds involving any A7 Network Sub-Agent. The wording covers the obvious case and one less obvious one: any transmittal from or to a Sub-Agent, and any transmittal from or to an account or a CVC address administered by or on behalf of a Sub-Agent. The rule asks only whether a named party, or an address that party administers, sits on either end.
FinCEN says compliance requires no tools or competencies beyond those covered financial institutions already use for their current AML/CFT programs.
The structural consequence is easy to miss. The operative test is membership on a list FinCEN maintains and updates, not a property of the transaction a filter can compute alone. An instruction not to transmit to addresses a party controls is enforceable only where those addresses appear somewhere a system can read. This is a predicate that cannot be machine-checked from the transaction alone, and the discussion of rule predicates that no filter can evaluate on its own frames it well.
There is a petition path, the human counterpart to the machine check. An entity that believes it was wrongly identified as a Sub-Agent may petition FinCEN for reconsideration, and if the challenge succeeds FinCEN would update the list. The list is the interface; the petition is the change request. Anyone who has maintained a denylist knows the property this creates: correctness depends on how quickly the list moves.
The Token, and What an Outside Reader Can Verify
The rule describes A7A5 as a ruble-backed stablecoin operating on the Tron and Ethereum blockchains, created by the A7 Network for Russian clients of OFAC-designated A7 LLC, a firm providing cross-border settlement platforms frequently used for sanctions evasion. It says the token is issued by Kyrgyzstan-based Old Vector LLC, which worked with Garantex, its successor exchange Grinex, and others in the token’s creation, issuance and trading. It says each coin is backed by ruble deposits at PSB, and draws the conclusion that every A7A5 transaction has a corresponding nexus to a sanctioned Russian bank.
Verification is where an engineer should slow down. Footnote 46 gives two contract addresses: an Ethereum address, 0x6fA0BE17e4beA2fCfA22ef89BF8ac9aab0AB0fc9, and a Tron address, TLeVfrdym8RoJreJ23dAGyfJDygRtiWKBZ, with pointer links to Etherscan and Tronscan. The rule separately cites an Etherscan contract page for the A7A5 token accessed 2026-08-17.
Calling the Ethereum contract at that address with standard ERC-20 view calls returned symbol A7A5, name A7A5, decimals 6, and a totalSupply of 581818931830000 raw units, equal at 6 decimals to 581,818,931.83 tokens. The Tron contract was not read. It confirms a contract exists at the cited address and answers the standard getters. It does not confirm the ruble backing, the reserve, who controls the contract, or whether the contract can pause or freeze balances, and the rule does not state those either. A verification is defined by what it rules out, and this one rules out a wrong address and little else.
Why the Definition Outweighs the Prohibition
The rule defines convertible virtual currency, or CVC, as a medium of exchange that either has an equivalent value as currency or acts as a substitute for currency, but lacks legal tender status. Up to that point the definition is ordinary. The next sentence is not. The rule says that despite A7A5 having legal tender status in at least one jurisdiction, for the purpose of this notice of proposed rulemaking the A7A5 stablecoin is included as a type of CVC.
A definition is the boundary of a machine: everything inside passes, everything outside is ignored. Most boundaries are a predicate a value either satisfies or does not, which any independent operator can evaluate the same way. Here the boundary is drawn by explicit inclusion of one named token. That is not a predicate the token passes; it is a decision the drafters wrote into the text, and the only way to learn it is to read the sentence.
This is why the definitional move matters more than the prohibition. A prohibition can be narrowed, redrafted, or withdrawn. A definition that carves out a special case tells a careful reader how the agency intends to classify tokens that resemble money but are not legal tender everywhere. The sentence does classification work a general test could not.
Five terms carry that load in the rule, and each one puts the check somewhere different. The third column below is the useful one: it says what an outside reader would have to consult to evaluate each term. For only one of the five, Sub-Agent, is the answer a list someone else maintains rather than a test the reader can run.
| Defined term in the rule | What the text says it means | What makes it checkable |
|---|---|---|
| Sub-Agent | Used with the identification mechanism and the list FinCEN maintains; an entity may petition for reconsideration of its inclusion, and if the challenge succeeds FinCEN would update the list. | Checkable only against the list FinCEN maintains; membership is not derivable from the transaction itself. |
| Convertible virtual currency (CVC) | A medium of exchange that either has an equivalent value as currency or acts as a substitute for currency, but lacks legal tender status; despite having legal tender status in at least one jurisdiction, A7A5 is included as a type of CVC for this notice of proposed rulemaking. | The general test is checkable in principle; the inclusion of A7A5 is not, because it is written as an explicit exception rather than as a test result. |
| Covered financial institution | Has the same meaning as financial institution in 31 CFR 1010.100(t). | Checkable by reference to that definition, which fixes the set of bound parties. |
| Transmittals of funds | The sending and receiving of funds, including CVC; for avoidance of doubt, that definition applies only to section 1010.668, and the definition in 31 CFR 1010.100(ddd) would not apply to section 1010.668. | Checkable against the rule’s own scope clause, which overrides the general definition for this section. |
| Recipient | The person to be paid by the recipient’s covered financial institution. | Checkable only once the covered financial institution on the receiving side is identified. |
The Claim About How the Token Was Used, and How Much Is Evidence
The rule makes a specific claim about usage. It says the A7 Network most often used A7A5 as a non-freezable bridging asset to convert into other, more widely accepted digital assets such as the stablecoin Tether (USDT), which may then be converted into the fiat currency of the customer’s choosing. To maintain liquidity, the rule says, the network uses its Sub-Agents or other trusted intermediaries such as digital asset exchanges, including over-the-counter brokers outside Russia in jurisdictions of concern, especially firms newly created or dramatically expanding their stablecoin trading operations, which could serve as liquidity providers.
That describes a flow, and tracing where money goes is what the rule delegates to analysts rather than asserting on its own authority. The write-up on tracing where funds actually move on a public chain is the relevant method, because the usage claim is only as strong as the on-chain analysis behind it.
The rule does not present that usage claim as original research. It cites third-party analyses, and every figure here belongs to those analysts, not to FinCEN or to this reading. It cites a Crystal Intelligence report by Hannah Curtis, dated 2026-07-30, whose headline claim is that one wallet now holds 94.5% of A7A5’s supply. It cites an Elliptic report dated 2026-07-29 on the fall of A7A5, a TRM Labs on-chain analysis dated 2026-06-12, and a Centre for Information Resilience report dated 2025-06-25.
None of those figures should be read as independently verified. The concentration figure is a claim by Crystal Intelligence, not a measurement performed for this article. The usage pattern is FinCEN’s characterization, supported by the analysts it cites. An engineer separates the normative text, what the rule would require, from the descriptive text, what the agency believes happened. This paragraph is descriptive, and it stays attributed.
The Precedent This Rule Sits Inside
The rule was issued under section 9714(a) of the Combating Russian Money Laundering Act, Public Law 116-283, as amended by section 6106(b) of the National Defense Authorization Act for Fiscal Year 2022, Public Law 117-81. Under that authority, FinCEN finds transactions involving any company operating outside the United States that is controlled by the A7 Network to be a class of transactions of primary money laundering concern in connection with Russian illicit finance.
FinCEN proposed a special measure against a whole class of transactions once before. On 2023-10-23 it published a proposal on convertible virtual currency mixing as a class of transactions, document 2023-23449, whose comment period closed 2024-01-22. A Federal Register API search on 2026-10-06 for FinCEN documents returned no final rule corresponding to that 2023 proposal. That is what the search returned and nothing more; it does not forecast this proposal’s fate.
What This Rule Cannot Do on Its Own
This is a proposed rule, not a final rule, and its operative test is membership on a list. Two limits follow. The first is procedural: while it remains a proposal, the obligation binds no one. The second matters for anyone who looks up the cited addresses. A test built on a list is only as current as the list, and an agency’s list moves on the agency’s schedule, not the chain’s.
The freeze question deserves an honest answer. The rule describes A7A5 as non-freezable in the context of how the network used it, as a bridging asset. Whether the Ethereum contract at the cited address has a blacklist or freeze function cannot be confirmed from the proposed rule text, because the rule does not say so, and confirming it would require auditing the contract source. For a stablecoin, the freeze switch is the most consequential access-control question in the design, and the discussion of who actually holds the freeze switch for a stablecoin matters here, precisely because the rule leaves the question open.
A further gap separates a prohibition from a chain. A prohibition on transmittals binds covered financial institutions, a set defined by 31 CFR 1010.100(t). That set is not the same as everyone who can move a token on a public chain. A rule can compel an institution to stop transmitting funds. It cannot, by its own force, stop a private key from signing a transfer. Reading the rule as though it were the second would be a category error.
What this reading accepts as settled is specific and checkable. A proposed rule exists as published text. That text names a token and two contract addresses. The Ethereum address answers the standard getters with the name, symbol, decimals and supply stated above. The cited analysts made the usage and concentration claims the rule repeats, and the rule attributes them. These hold regardless of what happens to the proposal.
The limits are equally specific. This reading does not establish the ruble backing, the reserve, who controls either contract, or whether a freeze or blacklist function exists at the cited Ethereum address. It does not establish that the rule becomes final, and it does not establish anything about the network’s conduct beyond what the rule and its cited sources state. The measure of an engineering read is the size of the gap it leaves open, and this one leaves the freeze switch, the reserve, and the rulemaking itself unresolved.
