Congress wanted a study of crypto mixers. The bill stalled, so we ran it on-chain.

01 · The askWhat Section 309 says

Most of the fight over the CLARITY Act was about who regulates tokens and what the president may own. Section 309 drew little notice. It is one page long and it orders a study.

It starts by saying what it means by a mixer. The words below are from the bill as the Senate Banking Committee reported it on June 1, 2026 (H.R. 3633, page 447):

"the term 'digital asset mixer and tumbler' means a smart contract, or set of smart contracts, that obfuscate or eliminate the source or other forms of identification of the holder of a digital asset, including by pooling assets from different holders and redistributing those assets among holders."

H.R. 3633, Section 309(a), as reported in the Senate, June 1, 2026

Then it gives the Treasury a deadline, "not later than 1 year after the date of enactment", and a list of six things the report must cover. These are the bill's words.

The committee's own summary, dated May 12, puts it in one line. The report must cover how mixers work, "how much and what share of their use is illicit versus legitimate", how exposed exchanges and banks are, how other countries treat them, and what new rules the Treasury wants.

02 · The voteA study nobody has to write

On September 15 the bill came to the Senate floor for a procedural vote and did not get the votes it needed. CoinDesk wrote that it "saw bipartisan opposition" and that "its future is now in limbo". The same report puts most of the blame on the ethics section and the coming election. Mixers were a side issue.

They came back into view on September 27, when Senator Cynthia Lummis, one of the bill's authors, blamed Democrats for the vote in a post:

"Democrats demanded a government study on crypto mixers and tumblers, the tools criminals use to launder digital assets. The Clarity Act mandates it. Democrats voted no on protecting consumers."

Senator Cynthia Lummis, on X, September 27, 2026

By her account Democrats asked for the study and Republicans wrote it in. Neither side got it. The list is still a fair one, and a mixer built from smart contracts is an odd thing to need a year to study. Every deposit and every payout is public. We took the six questions to the Ethereum record and answered what it can answer.

One note before the numbers. We quote the June text. Senators changed parts of the bill in the days before the vote, and we have not seen a later copy of this section.

03 · Question oneHow much goes through

The bill asks first for "historical transaction volume". We counted Tornado Cash, because it is the largest mixer that fits the bill's wording. You pay in a fixed amount, wait, and take the same amount out to a fresh wallet. Our seven-year audit from August covers how it works and what the 2022 sanction did to it. Here we only bring the count up to date.

Tornado Cash has four ETH pools, for 0.1, 1, 10 and 100 ETH. This is what went into them.

The last row is not yet a full quarter and it is already the second largest on the list. For the year so far the total is 565,259 ETH, in 58,037 deposits.

Two things about the shape. Most of the ETH moves through the largest pool, 100 ETH at a time. The smallest pool takes more deposits than any other, 43% of them. And more has come out this year than has gone in.

The ETH pools are only part of it. Tornado Cash also has pools for DAI, USDT and USDC. A full study would count those, and mixers on other chains, too.

04 · Question twoHow much of it is crime

This is the question the whole section is built around: the "percentage of transactions" used by "actors engaged in illicit finance". It is also where the chain runs out.

Start with what it does show. We added up what each wallet paid in this year.

The median is the wallet in the middle when you line them all up by size. Half paid in less than that, half more.

So Tornado Cash is two crowds in one place. Thousands of small wallets make up nearly half the users and almost none of the money. A hundred large ones make up half the money. Any figure for crime will turn on who those hundred are.

We looked. We ran the 60 largest through our address labels, which tag exchanges, hackers, scams and sanctioned wallets. None of the 60 came back with a label. About half of them paid in all their ETH within two hours.

That does not make them criminals, and it does not clear them. It means a percentage cannot be read off the chain. Someone has to tie each large wallet to a known theft, one by one, and the answer will only ever be a floor: the share that was caught.

Case work shows what that looks like. On September 27 a fake copy of Upbit's GIWA chain drew real ETH into a bridge on Ethereum. We traced it.

That is one theft, and it adds 514.9 ETH to the known-crime side of the ledger. A third wallet was still feeding the mixer when we stopped counting, and our report on the bridge follows it. The other way round happens too. After the Bitget hack several reports said the thief had pushed thousands of ETH through Tornado Cash. We traced every deposit in those hours back three steps and none led to the attacker. A study that took the headlines at their word would have counted that money as crime.

05 · Question threeWhere exchanges come in

The third question is about "centralized exchanges and traditional financial institutions", and how exposed they are. On-chain, that means two things. Does money reach a mixer from an exchange? And does money leaving a mixer end up at one?

We did not scan every wallet for this. We took three examples: the three wallets that drew the most from the 100 ETH pool between September 14 and 27. Then we followed each one forward. The second and third also drew from the smaller pools, and their figures count those.

None of the three went straight to a CEX. One kept its ETH and two went to bridges. Three wallets prove nothing about the rest. They do show why this one is hard to answer. By the time mixer money reaches an exchange, if it ever does, it has often changed coin or changed chain.

The other direction is easier to see. The two wallets that set up the fake GIWA bridge were first funded the day before, both from the same address. Our labels tag that address as the hot wallet of an instant-swap service. So a company with customers sat at the start of that trail, even though none sat at the end.

Whether exchanges are "adequately implementing" their checks is the second half of the question. That lives in their own records. No blockchain can show it.

06 · Questions four to sixWhat the chain cannot answer

Question four asks about "non-illicit uses… including privacy benefits". The chain offers one hint here and no more. There are 4,463 wallets that paid in 1 ETH or less. Small, steady use looks like people keeping a salary or a purchase private. But a hint is all it is. A small deposit can be a test run before a large one.

Questions five and six, on how other countries handle mixers and what the law should be, are for lawyers and lawmakers.

There is one more thing the bill's own wording leaves out. It defines a mixer as "a smart contract, or set of smart contracts". Tornado Cash fits. A mixing service run by a company on Bitcoin, which holds your coins and sends back other ones, is not a smart contract at all. Read as written, the study would cover the kind of mixer that is fully public and skip the kind that is hidden. We read the words as a layman would. A lawyer may read them another way.

07 · What it adds up toThree answers, three gaps

Congress gave the Treasury a year for this. The first question takes an afternoon. The second is the slow one, and it will stay slow for as long as the largest wallets have no names. If the study is ever ordered, that is where the year will go.

08 · MethodHow we counted

We counted native ETH paid into and out of Tornado Cash's four ETH pools on Ethereum, and into the router contract that most deposits pass through. The pool and router addresses are tagged as Tornado Cash in our labels. The window for this year's figures is January 1 to the end of September 27, 2026, UTC. Quarterly figures for 2025 cover full quarters. Our figure for the last quarter of 2025 matches the one in our August audit, which was built from a separate pull.

A wallet that paid in is the address that sent ETH to the router or straight to a pool. One person can use many wallets, so 9,738 is a count of addresses. The number of people is smaller. Payouts include the small fee that goes to the relayer who sends the withdrawal.

For the 60 largest wallets we looked up every label we hold. For the fake GIWA bridge and the three example wallets we followed transfers with Coinpath, one wallet at a time. We did not try to match any deposit to any withdrawal. Tornado Cash is built to make that impossible, and we did not attempt it.

If you need a trail like this followed for a case of your own, that is what our investigation team does.