Zero-Knowledge Proofs in Finance: Privacy, AML, and the Questions to Ask First

Updated November 10, 2026.

August 8, 2022. The US Treasury puts a piece of software on the sanctions list.

Not a person. Not a company. A set of smart contracts on Ethereum called Tornado Cash. Treasury says it laundered more than $7 billion since 2019, including over $455 million stolen by North Korea's Lazarus Group.

Tornado Cash runs on zero-knowledge proofs. You deposit ETH into a shared pool, and later you withdraw it to a fresh address with a proof that says "I'm owed money from this pool" without saying which deposit was yours. The math works. That's the problem. It works the same for a payroll department hiding salaries from competitors and for a state hacking group cashing out a heist.

If you run compliance at a bank, an asset manager, or an audit firm, Tornado Cash is the first thing you think of when someone pitches you on zero-knowledge. It should be. But it's not the whole story, and the rest of the story is where the useful stuff lives.

The problem was never the cryptography. It was a pool that let anyone in and showed no one anything.

Where Tornado Cash stands today

The legal status has moved a lot, so here's where it is as of November 2026.

On November 26, 2024, the Fifth Circuit ruled in Van Loon v. Treasury that Tornado Cash's immutable smart contracts aren't "property" that can be blocked under IEEPA. Treasury removed Tornado Cash from the sanctions list on March 21, 2025, framing it as an exercise of discretion rather than a concession.

The sanctions went away. The criminal cases didn't.

On August 6, 2025, a federal jury in New York convicted co-founder Roman Storm of conspiring to run an unlicensed money transmitting business. It deadlocked on the money laundering and sanctions counts. The retrial on those two counts is set for April 26, 2027, and as of late August 2026 his motion for acquittal was still undecided. In the Netherlands, developer Alexey Pertsev was convicted of money laundering in May 2024 and sentenced to 64 months. His appeal is pending.

There's one more piece. On August 21, 2025, Acting Assistant Attorney General Matthew Galeotti said DOJ won't approve new unlicensed money transmitting charges against developers of software that's truly decentralized and never takes custody of user funds. That's a prosecutorial policy, not a statute. It can change.

What does that mean for you? Two things. Privacy tools carry real legal risk for the people who build them. And "the code is neutral" hasn't held up as a defense when the tool does nothing to keep criminal money out.

Why institutions want privacy on-chain anyway

Here's the uncomfortable flip side. A public blockchain is a terrible place to run a treasury.

Every transfer, balance, and counterparty is visible to anyone with a block explorer. If your fund moves $50 million into a lending protocol, competitors can see it. So can the bots that watch the pending transaction queue and trade in front of large orders, which is what people mean by MEV. Your clients' positions show up too, and you owe those clients confidentiality.

That's the core enterprise problem, and the original version of this post got it right: public ledger transparency leaks strategy, positions, and client information. Banks don't operate that way off-chain, and they won't on-chain.

A zero-knowledge proof is the tool that gets you out of that bind. It lets one party prove a statement is true without revealing why it's true. If you want the mechanics, start with my plain-language explainer on zero-knowledge proofs and the zk-SNARKs deep dive. For finance, it shows up in four places.

Private transactions that don't become the next Tornado Cash

The design question is simple. Can you get transaction privacy and still keep criminal money out?

Three approaches are live.

Viewing keys. Zcash pioneered this. Every shielded address has a viewing key that grants read access without spend access. You can hand it to an auditor or regulator so they see your transactions while the public sees nothing. That's selective disclosure, and it maps neatly onto how audits already work, with two limits: incoming payments never reveal who sent them, and the spender's wallet can leave outgoing details out. So a viewing key supports an audit; it isn't a complete record of counterparties. The Zcash post covers it in depth.

Privacy Pools. In 2023, Vitalik Buterin, Chainalysis chief scientist Jacob Illum, and several academics proposed a fix for the Tornado Cash problem. When you withdraw, you prove your deposit belongs to an approved "association set" of deposits that excludes known bad funds. You stay private inside a clean crowd. A firm called 0xbow launched it on Ethereum mainnet in March 2025. It's early, and the open question is who decides what counts as clean.

Permissioned privacy layers. EY built Nightfall, an open-source zero-knowledge layer for private transfers between known business parties. The April 2025 version is a ZK rollup. Celo deployed it in September 2025, and StarkWare integrated it into Starknet in February 2026 for confidential institutional payments. The model is different from Tornado Cash at the root: participants are known and checked, and only the transaction details are hidden.

Europe has already drawn its line. The EU's Anti-Money Laundering Regulation (2024/1624) bars banks and crypto-asset service providers from keeping anonymous accounts and pushes them away from anonymity-enhancing coins, starting July 10, 2027. Anonymous privacy is on its way out of regulated finance in the EU. Accountable privacy, where someone with authority can still look, is the version that survives.

Proof of reserves: a real step, not an audit

November 2022. FTX collapses, and every exchange scrambles to prove it actually holds customer money.

The basic proof-of-reserves trick is a Merkle tree. The exchange hashes every customer balance into a tree and publishes the root, so you can check that your balance is included. The weakness is that a plain Merkle tree can hide negative balances, or leak information about other accounts.

Zero-knowledge proofs close that gap. In February 2023, Binance added zk-SNARKs to its proof of reserves, proving that no account is negative and that balances sum to the stated total without revealing any single one. OKX added zk-STARKs in April 2023.

That's a real improvement. It's also not an audit, and if you sign opinions for a living, this is the part to underline.

A ZK proof of reserves proves what's on the ledger. It says nothing about the loan the exchange took out last Tuesday that isn't in the tree. It covers one moment, so assets borrowed for the snapshot look fine. And it proves the math was done right on whatever data went in. Garbage in, perfectly verified garbage out.

The PCAOB said as much in a March 2023 investor advisory: proof-of-reserve reports aren't audits, they don't follow PCAOB standards, and they don't tell you the whole story about liabilities. Zero-knowledge makes the asset side more trustworthy. It doesn't touch the liability side.

KYC credentials: prove the check, keep the passport

This is the use case I'd watch closest.

Today, KYC means copying a customer's passport, utility bill, and birthday into your systems, then into the next firm's systems, then into the next. Every copy is breach exposure. A zero-knowledge credential flips that. A regulated issuer checks the customer once and signs a credential. Later, the customer proves a statement about it, such as "I passed KYC with a regulated institution," "I'm not on a sanctions list," or "I'm over 18," without handing over the documents.

It's not theoretical anymore. Google added zero-knowledge age proofs to Google Wallet in 2025 and open-sourced the libraries in July 2025. The European Commission declared its age verification app technically ready in April 2026, and every EU member state has to offer a digital identity wallet by the end of 2026.

The hard part isn't the cryptography. It's reliance. US customer identification rules only let a bank rely on another institution's checks under narrow conditions.

A ZK credential tells you a check happened. It doesn't tell you the check was any good, whether the credential was revoked yesterday, or who's liable if the issuer got it wrong. Those are contract and policy questions, and nobody has fully answered them yet.

The Travel Rule still applies, too. Proving a customer is clean doesn't excuse you from sending originator and beneficiary information where the rules require it.

What regulators have actually said

Less than the vendors imply. More than nothing.

• SEC. At the Crypto Task Force's December 15, 2025 roundtable on financial surveillance and privacy, Chairman Paul Atkins named zero-knowledge proofs, selective disclosure, and wallet design as privacy-preserving tools, and warned that crypto could otherwise turn into "a financial panopticon."
• Treasury. The GENIUS Act, signed in July 2025, required Treasury to study new ways to detect illicit finance. Its August 18, 2025 request for comment asked about digital identity verification built to protect privacy. Comments closed October 17, 2025.
• Central banks. The BIS Innovation Hub's Project Mandala, with the central banks of Australia, Korea, Malaysia, and Singapore, built a proof of concept that uses zero-knowledge proofs to run cross-border compliance checks before a payment settles, without exposing the underlying data.
• EU. The AMLR ban on anonymous accounts, above.

Notice what's missing. As far as I can find, no US regulator has said a zero-knowledge proof satisfies a specific AML, sanctions, or recordkeeping obligation. The signals are friendly. The rules haven't changed. Plan accordingly, and talk to counsel before you treat any of this as a compliance answer.

The questions to ask before you rely on it

When a vendor, a protocol, or your own innovation team brings you a zero-knowledge system, this is my list.

1. What exactly does the proof prove, and what does it leave out? Write the statement in one sentence. "Balances sum to X and none are negative" is not "the exchange is solvent."
2. Who wrote the circuit, and who audited it? A bug in the circuit can break everything. Zcash had a counterfeiting vulnerability in its original proof system that could have let someone mint coins undetected. It was found and quietly fixed in 2018, then disclosed in 2019.
3. Does it depend on a trusted setup? Some SNARKs need one, and if it's compromised, proofs can be forged. STARKs don't. The STARKs post covers that trade-off and the quantum angle.
4. Who can see what? Who holds viewing keys? Can your examiner get access on demand? What happens if a key is lost or leaked?
5. Who can change the system? Plenty of ZK systems still have admin keys or upgrade councils that can change the rules. That's a counterparty, and it belongs in your risk assessment.
6. For credentials, who issued it and how is it revoked? What's your legal basis for relying on it, and who's liable if it's wrong?
7. Can you still produce records? BSA recordkeeping doesn't care how elegant the proof is. If you can't hand an examiner the transaction history, you have a problem.

None of these are exotic. They're the questions you'd ask about any third-party control. Zero-knowledge just makes it easier to forget to ask them, because the math sounds so final.

A proof is only as trustworthy as the statement it proves and the people who wrote it. Start there, and zero-knowledge becomes what it should be for finance: a way to show your work to the people entitled to see it, and nobody else.