Supervisors learned to read the last crisis off bank balance sheets. Two IMF papers published this week argue the next one will not be there to read.
The IMF’s working paper on financial market infrastructure in a tokenised economy, released 1 July, and a companion blog by Financial Counsellor Tobias Adrian a day later, are not the familiar tokenisation pitch. The familiar pitch is that putting assets on a shared ledger makes settlement faster and cheaper. The IMF’s point is different, and less comfortable: faster settlement moves the risk.
The problem
Start with what atomic settlement actually does. Today, a securities trade runs through three separable stages — execution, clearing, settlement — each with its own institutions, timing buffers, and points where a human can stop the machine. Tokenisation collapses those stages into one software-governed step. The trade, the transfer of the asset, and the transfer of cash happen together, on the same ledger, enforced by code.
That is the efficiency everyone advertises. It is also the risk transfer nobody underwrote.
In the current system, the buffers are where risk lives — and where it is managed. A central counterparty stands between buyer and seller. A central securities depository keeps the record. Margin is called on a schedule. When something breaks, there is a queue, a phone call, a discretionary pause. The IMF’s language is precise: in tokenised systems, “execution, settlement, and aspects of risk management migrate toward shared infrastructures and programmable logic.” The risk does not disappear. It moves off the balance sheets of the banks and funds that used to carry it and onto the infrastructure — the ledgers, the smart contracts, the operators running the rails.
The analysis
Here is why that matters more than the settlement-speed story. Risk that lives in code propagates at the speed of code.
The IMF is blunt about the mechanics. Liquidity demands “could emerge instantly.” Collateral calls “could become fully automated.” Market stress could “spread faster than regulators or financial institutions can respond.” In a margin spiral today, automation is partial and human judgement is the circuit breaker — someone decides whether to widen a haircut, extend a deadline, or let a position fail. Wire that logic into self-executing contracts and the circuit breaker is gone. A collateral shortfall triggers a call, the call triggers a sale, the sale moves a price, the price move triggers the next call — with no interval in which a supervisor can intervene, because the interval is what the technology was built to remove.
Adrian names the endpoint. “Critical smart contracts,” he warns, “could become too important to fail.” That is a deliberate echo. The phrase that defined 2008 described institutions. He is applying it to code.
The companion working paper — authored by Yaiza Cabedo, Tommaso Mancini-Griffoli, Fabian Schär and Nicolas Zhang — is more cautious, and the caution sharpens the point rather than softening it. Tokenisation, it argues, will reconfigure financial market infrastructure, not eliminate it. Smart contracts absorb the routine work that central securities depositories, central counterparties and trade repositories do now: record-keeping, settlement, collateral management, reporting. But the functions requiring “legal certainty, governance, accountability, and discretion” stay institutional. The result is a hybrid: automation runs the routine, institutions handle the exceptions.
Read the two papers together and the fault line is obvious. The routine — the part that now runs at machine speed — is exactly where fast-propagating risk lives. The exceptions — the part where humans still decide — run at human speed. A crisis is a wave of exceptions arriving faster than exceptions can be processed. The hybrid model automates the propagation and leaves the intervention manual. That is not a hedge against the risk. It is the shape of the risk.
The implication
Which brings the supervision problem into focus. Regulators are organised around entities. They license banks, examine funds, supervise clearinghouses. Their entire apparatus assumes that the thing holding the risk is a legal person with a balance sheet, a charter, and a supervisor assigned to it.
Tokenisation moves the risk to something that is none of those things. A settlement protocol is not chartered. A smart contract has no supervisor. The operator that maintains a permissioned ledger may sit outside the perimeter that any single regulator can reach. The IMF’s own framing concedes the gap: supervisors must now oversee “not only institutions but also the software governing transactions.” That is not a tweak to existing mandates. It is a category the mandates do not contain.
Adrian’s proposed answer is coordination — “strong domestic policy frameworks remain the first line of defense,” he writes, “but international coordination is essential.” He is right, and the timeline is against him. Coordination among regulators is measured in years; the Global Layer One compliance framework published on 22 June — a joint effort whose contributors include the IMF itself, the Banque de France, the Monetary Authority of Singapore and JPMorgan’s Kinexys — is an early, partial example of how slow and bespoke that work is (Bitcoin.com News).
This is the part the settlement-efficiency framing hides. Every gain the IMF credits to tokenisation is real: less counterparty exposure between trade and settlement, continuous liquidity management, compliance embedded in the asset. But each of those gains is the same event described from the other side. Less counterparty exposure means the counterparty is no longer the shock absorber. Continuous settlement means no buffer to halt a cascade. Embedded compliance means the rules live in code that executes whether or not a human agrees with the outcome.
The IMF has not called tokenisation dangerous. It has called it structural — a change in where risk sits, not how much there is. That is the more serious claim. A dangerous product can be banned. A structural shift can only be supervised, and supervision is scoped to the institutions that are, by the IMF’s own account, ceasing to be where the risk lives.
The plumbing of the next crisis is being laid now. It is being laid in code. And the people whose job is to watch for it are still, mostly, reading balance sheets.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.