a16z crypto published Banks Don’t Need Closed Blockchains (written by Rebecca Rettig of Jito Labs). The legal reading is solid. The BSA and OFAC ask for control proportionate to identified risk, not zero risk, and no rule says a bank must own or screen the infrastructure its messages travel over. I have no argument with any of that.
HOWEVER, the conclusion does not follow. Banks are not sitting in permissioned networks because a statute traps them there. They are sitting there because that is where their trust model, their commercial agreements, investments, and their competitive advantage already live. I made this case in The Realities of On Chain Finance: Why Closed Ledgers Will Lead the Way and the last 18 months have only confirmed it.
Trust in banking is a commercial construct, not a consensus algorithm
Trust in banking rests on three things: commercial agreements that assign roles and liability, the ability to withstand regulatory scrutiny, and market scrutiny of your financial position and your capacity to assume risk. None of those three is a technical property. A validator set cannot supply any of them.
This is the point I have been making for fifteen years, in Building Networks and “Openness” (2011), Banks/Non-Banks and Commerce Networks (2014), Transformation of Commercial Networks (2016), Open Banking, Open Payments and Trust Networks (2020) and Governance in Payments (2025). As I put it in JPMorgan, Citi and TCH: Tokenized Deposits ON Chain: the chain is never the source of commercial trust, the agreements are.
Nameless validators are not a risk model
Technologists like to present trust-minimized consensus as a substitute for managing risk. It is a substitute for one narrow function, ordering transactions, and nothing else.
Institutions that have spent two hundred years building the capital, the capability and the supervision required to assume risk are not going to write off that investment to join a network where their peers have made no equivalent investment. That is not conservatism. It is the correct read of where the value sits. I covered the institutional version of this in Distributed Ledger Governance, where the interesting design is Canton with Visa as a Super Validator, a permissioned governance role layered onto a shared ledger. Governance and operational oversight, not technical specifications, are what drive institutional participation.
Why banks work in closed networks
Here is my suggestion for anyone proposing that open chains replace closed bank networks: spend a year working with regulators, then come back and tell me about your validator set. Twelve months of examinations, remediation letters, and explaining to a supervisor why a pseudorandom party in an undisclosed jurisdiction processed a client transaction will clarify the design problem faster than any whitepaper.
Banks are responsible for managing known interactions with counterparties. Every counterparty. Closed networks are built for known interaction, defined roles, defined permissions, defined recourse. Public blockchains were built for anonymous interaction. That is the whole design intent, and it is a feature for the market they were built for. It is the opposite of the requirement a bank carries.
The a16z paper answers this by analogy to the internet and the telephone network, neutral carriage, no selection, no dealing. The analogy holds legally and still loses commercially. Banks do not buy infrastructure on the basis of what they can defend in an enforcement action. They buy it on the basis of what it lets them promise a client.
Bank transactions are confidential, and public ledgers are no longer opaque
Client positions, counterparties and strategies are confidential. The paper is honest that the cryptography addressing this is partly aspirational: confidential transfers with an auditor key are shipping but thinly used at institutional scale, and proofs of unsanctioned status and provenance remain in pilot and research.
There is a second problem, and it runs the other way. Governments have largely succeeded in removing the anonymity from public blockchain transactions. That capability is how crypto fraud schemes and AML end runs get found, and it is a good thing. It also means a bank putting flows on a public chain is publishing a permanent, analyzable transaction graph to every party that has invested in chain analytics, including competitors and foreign states. Privacy is not a solved objection. It is the objection.
Consortium is the mechanism, and it is already running
Banks act in consortium to manage counterparty risk and to reuse operating investments they have already made. This is what I described in Chain of Trust in 2022: adoption starts in closed or bilateral networks of collaborators, because that is where participants can construct a competitive advantage and absorb the downstream effects on ancillary products.
Banks have also built the most successful commercial networks in history. Visa and Mastercard, plus the clearing and settlement utilities in financial markets. Those networks work because roles, standards, counterparties, trust and risk carry a shared and enforceable definition, the argument I made in A2A, FedNow and Instant Payments, Threat to V/MA? Nope and in Winning in Network of Networks. Set against that, an open chain offers a bank no financial incentive and carries real regulatory risk. ZelleUSD and the TCH tokenized deposit consortium are what the answer looks like when banks actually build it.
The steam engine pattern
The best summary I can give is that banks are using blockchain to create competitive advantage. That is the dynamic behind every innovation since the steam engine. New technology is adopted first by incumbents in established markets to gain an edge, and only later, if ever, does it produce a parallel market structure.
I understand why the open community would prefer to operate at parity with banks. It will not happen through a legal argument that the parity is permitted. Value exchange requires a commercial construct, a contract, a marketplace agreement, or a network with operating rules, which is the case I made in Discount “On Chain”. Banks will not move without a defined opportunity, and permission is not an opportunity. The same mistake sits inside eIDAS, where trust was assumed to be transferable by regulation rather than earned with capital and liability (Why eIDAS Will Fail in Banking).
How to read the headlines
Look suspiciously at any fintech article claiming banks are going “on chain.” When the system is closed, it is new technology inside an existing structure, making that structure more efficient. That is the entire story, and it is a big one. JPMorgan’s Kinexys is the best example: every participant vetted before joining, every party with a defined role and defined permissions, DLT doing the settlement work while the commercial construct does the trust work. I said the same about the SWIFT tokenized deposit announcement, which was a me too.
Verdict
The a16z paper proves banks may use permissionless networks. It does not establish that they should, because the reason they do not is commercial, not legal. Closed chains work best for banks because banks sell risk management, and risk management requires known counterparties, confidential flows, enforceable agreements, and a supervisor who can be satisfied. Closed chains deliver all four today. Permissionless chains deliver none of them, and the cryptography that might, one day, deliver two of them is in pilot.
Nobody is going to trade a two hundred year investment in assuming risk for a legal opinion that it was never strictly required.