A crypto custody rule is only as good as the day the custodian fails

The only question a custody rule answers

A custody rule is not written for the good days. It is written for the morning a custodian stops answering the phone. Every other requirement in it — the account statements, the independent verification, the segregation of client property — exists so that on that morning somebody can say, with evidence, which assets belong to which client and who is permitted to move them. If the rule cannot answer that question, nothing else it contains is worth much.

The Securities and Exchange Commission, the American agency that polices markets and the advisers who manage other people’s money, has proposed a dedicated legal framework for crypto custody in the advisory and wealth management market. The direction is sensible. The existing architecture grew up around securities, where ownership is a book entry at a depository and “taking” an asset means persuading an institution to change a line in a ledger. That architecture has been asked for years to stretch over something it was never shaped for.

Why the old answer does not transfer

Crypto assets break the analogy in one specific way, and the whole rule turns on it. A security is held. A crypto asset is signed for. Control rests on possession of key material, and key material can be copied without being removed. A custodian can hold a client’s coins and so can a departed engineer, a hardware vendor, a sub-custodian in another jurisdiction, or whoever wrote the recovery procedure — all at the same time, all invisibly, until one of them uses it.

That is why the familiar comfort words do not do the work here. Segregation, in the securities sense, means client property is kept apart from the firm’s own. In crypto it can mean a separate address while the signing authority sits in exactly the same place it always did. Independent verification can confirm that an address contains coins without establishing who is able to spend them. Proof of reserves is not proof of control, and only one of those matters in a failure.

Three things the text has to pin down

First, an exclusive control test that can be examined. Not a declaration that the custodian holds the assets, but a requirement to identify every party holding a key, a key share, or a recovery mechanism, and to have that attested on a stated schedule by someone who does not work for the custodian. A control standard that cannot be audited is a marketing claim.

Second, insolvency. The question of whether client crypto is the client’s property or part of the custodian’s estate should be answered in the rule, not discovered by a judge three years later while the assets sit frozen. Advisers placing client assets somewhere are entitled to know which outcome they are choosing.

Third, loss allocation. Keys leak. Hot wallets get drained. The rule should require that the contract between custodian and client states, before anything goes wrong, who bears a loss arising from compromise of key material and what capital or insurance stands behind that promise. Leaving it to negotiation means it will be resolved by whoever has the better lawyers at the worst possible moment.

The case against this argument

The strongest objection is that the Commission has been criticised for years — fairly — for regulating this asset class through enforcement rather than rulemaking, leaving advisers to choose between avoiding crypto entirely and relying on arrangements designed for something else. A dedicated framework, even an imperfect one, replaces guesswork with a path. The technical detail, the argument runs, is better handled in guidance and examination priorities, which can move as the technology does, rather than frozen into a rule that will be obsolete before it bites.

There is force in that, and the flexibility point is real. But guidance is not what a bankruptcy trustee reads. Examination findings arrive years after the fact and are rarely public. The documents that decide who gets the coins are the statute, the rule and the custody agreement — and if the rule is silent on control, insolvency and loss, those gaps are filled by whatever the custodian drafted for itself. A federal designation that signals safety while leaving the safety undefined does not reduce risk, it relocates it onto clients who reasonably assume the label means something.

There is also a timing point worth making. The Federal Reserve has just extended the comment window on its own insider-lending proposal to give people longer to analyse it. Crypto custody is more technical than insider lending, not less. A short consultation on a rule this engineering-heavy will produce comments from the firms with standing counsel and almost nobody else.

What would settle it

Everything above is written against a proposal whose text decides the matter. The Commission can close this argument in a paragraph by publishing a control standard with an audit requirement attached, and by stating the insolvency treatment on the face of the rule. If it does, the remaining debate is about calibration — how often, how independent, how much capital — which is a debate worth having and a far better place to be than the one the market is in now.