The frozen account
A decision that can strand someone with no reason and no route back is the sharpest test of accountability there is.
There is a message that arrives with no warning and no author. Your account is under review. No timeframe is offered, no reason is given, and the name at the bottom belongs to no one you can reach. In the interval — which may last hours or may last months — the money is still yours in principle and unreachable in fact, the login still works but does nothing, and the person on the other end of the support line, if there is one, can tell you only that a review is under way. You have been found suspect by a system you cannot see, on grounds it will not state, and the burden of proving yourself innocent has been placed on you without anyone telling you what you are supposed to have done.
This is among the most consequential automated decisions ordinary people encounter, and it is almost invisible as a decision at all. A freeze does not announce itself as a judgment. It presents as a pause, a precaution, a temporary measure taken while something is confirmed. But a pause with no stated end and no stated cause is not a precaution; it is a verdict that has learned to describe itself as a delay. The person affected is stranded, and the institution that stranded them frequently cannot say, even internally, exactly why. A threshold was crossed. A pattern matched a pattern. The account was flagged, and the flag has consequences long before anyone examines whether the flag was right.
The trapdoor and the door
What makes the frozen account so sharp a test is the asymmetry it creates. The institution incurs almost nothing by freezing. The cost of a wrongful freeze does not land on the system that imposed it; it lands entirely on the person locked out, who may miss rent, payroll, or a medical bill while an opaque process runs its course. Because the cost is externalized, there is little internal pressure to make the freeze reversible, to state its basis, or to give the affected party a route back. The path in is instantaneous and automatic. The path out, if it exists, is manual, slow, and discretionary — and discretion exercised in the dark is indistinguishable from no discretion at all.
A door you can be pushed through but not walk back out of is not a door. It is a trapdoor, and calling it a review does not change what it is.
The difference between a door and a trapdoor is entirely a matter of the record. A freeze that comes with a record — of what triggered it, on what basis, against which rule, and by what process it can be lifted — is a decision that can be contested, corrected, and closed. A freeze that comes with nothing is a trapdoor: the person falls through, and whether they climb back out depends on luck, persistence, and whoever happens to pick up their case. The same action, imposed by the same system, is due process in one instance and a disappearance in the other, and the only thing that distinguishes them is whether anyone wrote down why.
What the affected person is owed
Standing is the right to be treated as a party to a decision that concerns you rather than as its object. The frozen account is where standing either exists or is exposed as fiction. A person locked out of their own funds is a party to that decision by any coherent definition, and yet the design of most freezes treats them as a passive subject of a process happening at them, not with them. To have standing here means, at minimum, to be told that a decision has been made, to be told the basis on which it was made, and to be given a place where an objection can land and be answered. None of those requires the institution to reveal a detection method or tip off an actual bad actor. They require only that the institution keep, and be willing to disclose to the affected party, the record of its own reasoning.
The objection to this is always the same. If we tell people why they were flagged, we teach the guilty how to evade us. But this confuses two different disclosures. There is the general shape of a detection system, which an institution may legitimately keep confidential, and there is the specific, individual basis for a specific freeze imposed on a specific person, which that person has every right to know. A bank need not publish its fraud model to tell a customer which transaction tripped the alarm. The claim that any explanation aids evasion is, more often than not, a way of avoiding the discovery that the freeze rested on very little — that the pattern it matched was a coincidence, that the threshold was set for the institution's convenience rather than the customer's protection, that no human ever looked.
The point of a record here is not to slow the freeze down. Speed can be legitimate; a genuinely suspicious pattern may warrant an immediate hold. The point is that speed and finality are different things, and a fast decision that cannot be explained or reversed has quietly become final without ever being examined. A record separates the two. It lets the freeze be immediate and lets the review be real — because there is now something to review, a stated basis to test against the facts, a rule to check the action against, and a lifting condition that turns an open-ended limbo into a question with an answer. Reference the standard set out in Admissible Reality: a decision that affects a person is admissible only if the person can contest it, and it can be contested only if it left a record they can read.
A frozen account is the moment an institution's account of itself is most needed and most often absent. The person on the other side is not asking for their transaction to be approved; they are asking to be told, in terms they can answer, why they were shut out and how they get back in. An institution that can answer that question has imposed a hardship, which is sometimes justified. An institution that cannot has built a trapdoor and is hoping no one asks where it leads. The record is the difference, and the person falling through it is the one who should get to hold it.
— Dispatches · Summit Cognitive
Continue from here
Turn the argument into a practice.
Get new dispatches, assess how your organization handles consequential decisions, or explore Summit Cognitive.