The ledger holds. Everything around it does not

In February 2024 a cryptocurrency wallet that had sat untouched for five years sent 8.2 bitcoin to an exchange account. The judge who later described that transaction at Southwark Crown Court put it plainly: the defendant had presumably assumed that after five years nobody was watching. The police were. That single movement led to two arrests in York and Leeds on 22 April 2024, and on 11 November 2025 to sentences of eleven years and eight months for Zhimin Qian and four years and eleven months for Seng Hok Ling.

The figures around the case are large enough to obscure the point. The Metropolitan Police seized 61,000 bitcoin, the largest cryptocurrency seizure ever made in Britain. The Crown Prosecution Service put the current value of the seized assets at around £4.8 billion. More than 128,000 people in China had invested in the underlying scheme.

Read the sentencing remarks and something else stands out. Not one thing went wrong with the blockchain. The ledger recorded every transfer, permanently, in public, and that is exactly why an investigation begun in 2018 could still reconstruct the movements six years later. Every failure in the case happened somewhere adjacent: at an exchange account opened in an associate’s name, on a laptop carried out of the country and later found in a safety deposit box, in a solicitor’s client account frozen by court order, in lies told to professionals about the source of funds, and finally in the seed words for one of the wallets, which police obtained when a defendant handed over the PIN to his phone.

That is the shape of almost every real incident in this field. What follows is a map of the five places the failures cluster. It is a map of consequences and defences, not of methods, and it is deliberately written so that nobody learns how to do any of it.

One: the consensus layer, which is the part that mostly holds

The attack the phrase “hacking the blockchain” conjures, rewriting settled history by outvoting the network, is the rarest thing on this list, and it is rare for an economic reason rather than a cryptographic one.

Security on a public chain is bought. On a proof-of-work chain it is bought with electricity and hardware; on a proof-of-stake chain it is bought with capital that can be destroyed if you misbehave. The cost of overwhelming the network scales with how much of that the honest participants have already spent. On a large chain the sum required is comparable to the market value of what you would be stealing, which is a poor trade even before you consider that the theft would be visible to everyone in real time.

The corollary is the useful part, and it is routinely ignored by organisations choosing a platform. The security of a chain is not a property of the software. It is a property of the budget defending it. A small chain running the same code as a large one, with a fraction of the participants, is a fundamentally different risk, and the risk falls as the price of attacking it rises. Anyone evaluating a ledger for a business purpose should be asking what it costs an outsider to gain a controlling share of the validation, and whether that number is rising or falling. Almost nobody asks. They ask about transactions per second.

Two: custody, which is a bank without the parts that make a bank safe

Most people who own cryptoassets do not hold them. An exchange or custodian holds them, and the customer holds a database entry saying they are owed some. That arrangement is a bank. It is not, in most jurisdictions and for most of the last decade, regulated like one.

Britain’s answer so far has been an anti-money-laundering registration regime administered by the Financial Conduct Authority, which is a control on who may operate rather than a guarantee about what happens to customer assets when a firm fails. The Qian case shows the same layer from the other side. Investor money went into an account opened at a cryptocurrency exchange under somebody else’s name, and was converted to bitcoin from there. No cryptography was defeated. An account-opening process was.

For a UK organisation the practical question is not whether an exchange has been hacked. It is what happens to your position if it is, or if it simply stops answering the phone. Ask who the legal owner of the assets is under the custodian’s own terms, whether client assets are segregated, in which jurisdiction an insolvency would be heard, and what the withdrawal path looks like on a day when everybody wants one at once. Those are the questions a treasurer would ask of any counterparty. They tend to go unasked here because the technology sounds novel enough to suspend ordinary diligence.

Three: keys, which are not a control on the system but the system itself

In conventional IT, losing a password is an inconvenience because somebody can reset it. There is no reset here. Whoever controls the private key is, as far as the network is concerned, the owner. There is no appeal, no chargeback and no helpdesk.

That has two consequences that organisations consistently underestimate. The first is that key management stops being an IT hygiene task and becomes the entire security programme. The second is that key loss and key theft are the same event from the network’s point of view, which means your recovery plan and your incident plan are one document.

The Qian case supplies the illustration. Police recovered a laptop from a safety deposit box containing 4,741 bitcoin, worth about £25.2 million at the time, while the balance of an original holding of more than 70,000 had already been moved elsewhere. Later, a defendant’s phone PIN gave investigators the seed words for one of the wallets. A seed phrase is not a credential attached to an account. It is the asset. Anything that can read it, including a photograph, a note-taking app, a password manager backup or a person, is part of the custody arrangement whether or not anyone drew it on the architecture diagram.

The defences here are old and dull: keys generated and held in hardware that never exports them, a genuine threshold scheme so that no single person or single device can move value alone, a written and rehearsed procedure for what happens when the person who ran the key ceremony leaves, and an inventory of every place a recovery phrase has ever been written down. None of this is exotic. Most of it is not done.

Four: the joints between chains

The most expensive category of loss in this field, measured by value per incident, has consistently been the connective tissue. Moving an asset from one chain to another is not a native operation. Something has to hold the original while an equivalent is issued elsewhere, and that something is a piece of ordinary software holding an extraordinary balance, typically controlled by a set of signing keys with a threshold: a defined number of signatures out of a defined total is enough to authorise a movement.

Two design facts follow. The security of the whole pool reduces to the security of the smallest sufficient set of those keys. And the people holding those keys are employees, who read email.

That is precisely the population targeted in the joint advisory issued on 18 April 2022 by the FBI, CISA and the US Treasury on North Korean state-sponsored activity against blockchain companies. The advisory carries the product identifier AA22-108A and its original PDF is dated 18 April; the web page shows only the last revised date of 20 April, which is the date most citations of it repeat. The advisory describes spearphishing campaigns aimed at employees of cryptocurrency firms, “often working in system administration or software development/IT operations”, using messages that “mimic a recruitment effort and offer high-paying jobs” in order to reach the victim’s machine and, from there, “steal private keys or exploit other security gaps”. A job offer, a laptop, a key. The blockchain in the middle is a bystander.

If your organisation is holding assets behind a signing threshold, the number that matters is not the total number of signers. It is how many independent organisations, on independent hardware, in independent locations, would have to be compromised on the same day. If the honest answer is one, the threshold is decoration.

Five: the person, which is where the money actually goes

Strip out the technology and the largest category of loss is fraud of a kind that would be recognisable to a Victorian magistrate. City of London Police reported that 34,673 people reported investment fraud in 2025, a rise of 31% on the previous year, and put the daily loss at £2.4 million. Cryptoassets are the wrapper for a large share of it, because they combine an investment story that most victims cannot evaluate with a settlement system that does not reverse.

The Qian scheme itself was not a technical exploit at any point. It was an investment fraud that ran for three years, took money from more than 128,000 people, and only then touched a blockchain, at the moment the proceeds needed moving. The technology was the laundry, not the crime.

The defence, for an organisation rather than an individual, is the same procedural control that defends against invoice fraud: no irreversible transfer on the strength of a message, however well written, without out-of-band confirmation against a record held somewhere the requester cannot edit. The reason that control matters more here is the irreversibility. A mistaken bank payment has a recall process, however painful. A mistaken on-chain payment has a public record of your mistake.

Immutability is a cost as well as a feature

The word does a lot of unexamined work in board papers. Immutable means your errors are also permanent. A wrong address, a duplicated payment, a contract deployed with a defect and a key written to the wrong place are all final in the same way that the good transactions are final. Organisations adopting distributed ledgers for their tamper resistance rarely price in the operational discipline that a system with no undo actually demands, and that discipline is where the running cost sits.

The related trap is a governance one. A great deal of what is described as decentralised has a small group of people who can upgrade the code, pause the system or move the treasury. That is not a criticism, because somebody usually needs to be able to do those things. It is an observation that the trust you removed from the ledger has reappeared as a list of names, and that list, not the consensus algorithm, is your real threat model.

What to ask before you sign anything

Three questions separate an organisation that has thought about this from one that has been sold to. Who can move value, and how many separate people and separate devices would have to fail for that to happen once? What is the recovery position if the key holder, the custodian or the vendor disappears tomorrow? And what breaks if the chain you have chosen halts for a day, which is a far more common event than a theft?

None of those are cryptographic questions. That is the finding. The technical literature on this subject is excellent and the incidents almost never turn on it. They turn on account opening, employment, procurement and the location of a piece of paper. Our wider coverage of that pattern sits under cybersecurity. The same displacement, where compromise arrives through a party you signed a contract with rather than through anything you own, runs through the supplier route into British public services.

The unhackable ledger is, for practical purposes, unhackable. It is also the only part of the arrangement that is.