Before You Stock the Shelf: Where Controls Belong in Tokenized Banking
Tokenized money is leaving the pilot stage. Stablecoins are settling real commercial payments, funds are moving on-chain, and banks are standing up tokenized deposits. According to Fireblocks analysis highlighted in a recent Cosmos article only 3.4% of the world's 290 largest banks had live tokenized deposit capabilities as of mid-2026, a share expected to reach about 21% a year later.
I've been making the case to bankers that this is a product opportunity. The bank that builds the rail is building a shelf, and the shelf fills up with things banks can't offer today. A treasury account that sweeps into a tokenized fund intraday. A CD a commercial client can post as collateral in minutes.
There's a question that comes before any of that, and it's the one bank risk teams ask me most. What happens when a payment goes wrong?
Every rail banks use today has an answer. ACH has returns. A wire can be recalled. FedNow has a request for return. On-chain money moves around the clock and settles in seconds, and on most networks there's no way back. Before a bank stocks the shelf, it needs to know where the controls live.
Stablecoins need controls inside the transaction
Stablecoins are where most banks will start, and they can run on both permissioned and permissionless chains. In permissionless environments, anyone can hold a wallet and anyone can receive a payment. That openness is why stablecoins move so easily between exchanges, fintechs, and businesses. It's also why the risk looks different from anything a bank runs today.
On a permissionless chain, the bank often doesn't know who's on the other end, and once a transfer settles, it's final. A mistyped address, a spoofed invoice, or a payment to a sanctioned wallet can't be pulled back. An issuer can freeze tokens in some cases, but a freeze is a blunt tool. It's centralized, it's slow, and it won't help a bank resolve an ordinary dispute with its own customer.
The instinct is to fix this at the chain. Permission the network, add allowlists, and let the protocol decide who can transact. I think that's the wrong answer. It sacrifices the openness that makes stablecoins useful and asks the base layer to make compliance decisions it was never built to make.
Controls belong at the point of regulated contact. That means, within the transaction, it's enforced by the bank or fintech that's accountable for it. Reversibility is the clearest example. With a time-delay escrow, funds sit in a smart contract for a short window before they can be released. If something is wrong, the payment can be recalled before it settles. The chain stays open, and the bank gets the same recall path it already has on every other rail.
What building it taught us
We've had these controls in the market since July, and a few lessons surprised us.
Verification comes first. Most bad payments are wrong before they're sent, so the cheapest control is confirming the receiver. Screening has to happen before release, because after settlement there's nothing left to stop. Release conditions need to be explicit and programmable, whether that's a time window, a second approval, or a confirmation from the receiving party. Multi-party confirmation mattered more than we expected. Businesses want two people to sign off on a large payment, the same way they do with wires today.
The biggest lesson was about who decides. When an inbound payment is flagged, the bank chooses whether to hold it for review or return it to the sender, under its own procedures. Coinbax’s job is to make both paths possible and auditable. The regulated party stays in charge of the outcome.
I've spent two decades selling to banks and credit unions and partnering with the companies that serve them. That experience is how our team at Coinbax knows how banks buy. They adopt something new when the controls they already expect come with it. Treat the regulated party as your primary user and the adoption conversation gets a lot shorter.
Tokenized deposits change the job
Tokenized deposits work differently, and the difference matters for how banks plan.
A tokenized deposit is a bank liability recorded on a programmable ledger. It usually runs on a network where every participant is known. The bank onboarded the customer, and the counterparties are other members of the network. The issuing bank keeps authority over its own ledger and can correct errors under the account agreements it already has.
That means the network absorbs some of the risk that controls have to carry on a permissionless chain. Not all of it. Authorized push payment fraud still happens when the customer is real, and the instruction is real. Money still has to cross into stablecoins and public chains when customers want it there, and those edges need the same controls. Inside the network, though, the job of the controls layer changes. It moves from protecting the payment to programming it.
This is where the shelf fills up. A treasury account that sweeps into a tokenized fund in the morning and back before close. A CD a commercial client posts as collateral in minutes. A real estate closing where funds release when the title company confirms. A supplier payment that settles when the goods are received. Each one is a product, and each one earns fee income banks don't earn today.
The same engine that screens and holds a stablecoin payment runs these conditions. A bank shouldn't need one system for risk on public chains and another for products on its own network.
Why the foundation matters
None of this works if the ledger underneath forces a bank to choose between openness and control.
Cosmos is built on the idea that a network can set its own rules and still connect to everyone else. A bank, or a group of banks, can run a network with known participants and its own governance, then connect outward when money needs to move. The ledger stays neutral, and the rules live with the institution that's accountable for them.
That's the same principle we built our controls on. Keep the base layer neutral and put accountability in the transaction. It's why we joined the Cosmos Partner Network when it launched in September, alongside the custody, compliance, and infrastructure providers banks need to accelerate their tokenized deposits and digital asset initiatives. The Cosmos Tokenization Suite gives banks the ledger and issuance technology. Our role in the network is the controls layer that lets a bank put real products on it.
Stablecoins now, tokenized deposits next
Bankers ask me whether they should start with stablecoins or tokenized deposits. The answer is both, in that order.
Start with stablecoins now. Customers already hold them, and a bank that can't accept them is losing deposits it will never see leave. Put reversibility and screening inside the transaction from day one, because on a permissionless chain that's where the risk lives.
Then build toward tokenized deposits knowing the job changes. The network carries more of the risk, and the controls do more of the work of building products.
When banks ask me when they need to start, I point them to the regulatory calendar. The GENIUS Act takes effect in January 2027, and bank implementations take six to nine months. The architecture decisions are being made now.
The tools to build something new are available today. The ledger is ready, the partners are ready, and the controls that make it safe are already in the market. The banks that move first will decide what these rails become. Build the shelf, put the controls underneath it, and stock it with something your customers have never been able to get from a bank.
About the Author
Peter Glyman is Founder and CEO of Coinbax, a member of the Cosmos Partner Network, and a financial technology executive focused on stablecoin payments and programmable money. He co-founded Geezeo, acquired by Jack Henry in 2019, and later served as Managing Director of Corporate Strategy at Jack Henry, leading initiatives across embedded payments, fintech partnerships, and crypto strategy. Today, his work focuses on financial infrastructure and digital assets, including stablecoins, smart contracts, and programmable escrow.
