Last Updated on August 8, 2026 by Karl Thompson
This article is about a sound. The two-tone beep of a card terminal is one of the most frequently heard noises in British life, and almost nobody knows what it means. It does not mean money has moved. It means a stranger’s computer, several hundred miles away, has decided you are good for it.
This is part of How Modern Society Works, a series on the hidden systems that shape everyday life. Each article takes something ordinary and works outwards to the system that produced it.
The previous chapters examined infrastructures that occupy land, absorb capital, hold ownership or distribute time. This one examines an infrastructure made of something less tangible and more consequential: trust. It asks how that trust is manufactured, who is inside it and who is not, and — the question that turns out to matter most — who pays when it fails.
What Happens When You Tap Your Card?
You tap a card for a £2.90 coffee. Under two seconds later, the terminal beeps and you leave.
What happened in those two seconds was not a payment. It was a conversation.
The terminal read your card and sent an encrypted request to the merchant’s acquiring bank. That request was routed through a card network — Visa or Mastercard — which is not a bank and does not hold anybody’s money, but operates the messaging system and the rulebook that lets thousands of financial institutions speak to one another in a common format. The request reached your own bank, the issuer, which ran a series of automated checks in milliseconds: is the card genuine, is it reported stolen, are there funds, does this purchase fit the pattern of your previous behaviour or does it look like somebody else holding your card in a city you have never visited?
Your bank then sent back one word, in effect: yes.
The beep is that word. It is an authorisation — a promise by your bank to the merchant’s bank that the money is good. The actual movement of funds happens later, in a separate process called clearing and settlement, in batches, hours or days afterwards, between institutions you will never deal with.
So the coffee shop gave you a coffee not because it received money but because it received a promise, from an institution it has no relationship with, transmitted through a network it does not own, according to rules it did not write. And you handed over a piece of plastic on the understanding that the same arrangement would hold in reverse.
This is an extraordinary amount of confidence to invest in a two-second interaction with a stranger, and it is repeated, in Britain, tens of billions of times a year.
What Are Bacs, Faster Payments and CHAPS?
Britain does not have one payment system. It has several, each built for a different combination of speed, cost and certainty, and most people use three or four a month without being able to name any of them.
Bacs handles the routine, high-volume, pre-scheduled work: salaries, pensions, Direct Debits. It runs on a three-day cycle that has barely changed in half a century, because for a bill that everyone knows is coming, cheapness matters more than speed.
The Faster Payments System handles the bank transfer you make from your phone, arriving in seconds. It was introduced in 2008 and quietly transformed everyday financial life — splitting a restaurant bill, paying a tradesperson, sending money to a child at university — by removing the three-day wait.
CHAPS, operated by the Bank of England, handles high-value, same-day, irrevocable payments: house purchases, financial market settlement, the movement of very large sums where certainty is worth paying for.
SWIFT, a Belgium-based cooperative, is the messaging network through which banks worldwide send each other secure payment instructions, connecting more than 11,000 institutions.
Ownership varies in ways that matter. Visa and Mastercard are large American public companies. Pay.UK — which operates Bacs and Faster Payments — is a not-for-profit, and is possibly the most important organisation in Britain that almost nobody has heard of. CHAPS is run by the central bank. Oversight sits with the Bank of England, the Financial Conduct Authority and the Payment Systems Regulator.
The structure is worth noticing because it is unusual among the infrastructures in this series. The previous chapters described systems concentrated in a very small number of private hands. Britain’s domestic payment rails are a genuine mixed economy: some public, some mutual, some commercial, with the card layer dominated by two American firms sitting on top of a domestic infrastructure that is not.
Abstract Systems
Anthony Giddens gave modern societies a useful description of themselves when he argued that they run on abstract systems — mechanisms that allow people to coordinate across time and space with strangers they will never meet and processes they could never explain.
He identifies two. One is expert systems: bodies of technical knowledge that organise large areas of our environment and that nobody involved could personally verify. The other is symbolic tokens — media of exchange that circulate without regard to the particular people handling them. His main example of the second is money, which is why payments are not merely one illustration of his argument but close to its original case.
His point was not simply that these systems exist, but that living in modernity means continually placing trust in them without the kind of evidence that trust ordinarily requires. You cannot inspect the aircraft’s maintenance record, audit the water treatment plant, or verify the fraud detection algorithm. You proceed anyway, because the alternative is not proceeding at all. Giddens called the resulting stance a kind of routine bracketing-out of risk — not ignorance exactly, but a practical willingness not to ask.
Payments are close to the purest example available. Consider what you actually know when you tap a card. Not how the encryption works. Not which company operates the terminal. Not where the acquiring bank is. Not what happens if the transaction is duplicated. Not what recourse exists if it goes wrong. You know that it has worked before.
That is the entire evidential basis, and it is enough, and it is enough for nearly everyone nearly all of the time.
But Giddens’ account has a limitation that this article is going to press on. It describes trust as something individuals extend to systems, and treats the extension as more or less universal — a general condition of modern life. What it says less about is that the trust runs in two directions, and that the return direction is neither universal nor free.
The system trusts you back only conditionally. It assesses you, prices you, and can decline you. And when the trust breaks down in either direction, somebody bears the loss — and who that somebody is turns out to be a political decision that has been made, remade, and fought over.
How Often Do UK Bank IT Systems Fail?
Start with the simplest failure: the infrastructure stops working.
On Friday 31 January 2025, Barclays suffered a three-day IT failure. Around 5% of attempts to log into its mobile app and online banking failed outright, and of customers who did get in and tried to make a payment, 56% of those payments failed. The bank later attributed it to a software problem in a critical module of its UK mainframe operating system.
The timing could hardly have been worse. 31 January is the deadline for self-assessment tax payments and, for a very large number of people, payday.
The Treasury Committee then asked a broader question, and the answer is the most important single fact in this article. Writing to nine major banks and building societies — Barclays, HSBC, Lloyds, Nationwide, Santander, NatWest, Danske, Bank of Ireland and Allied Irish Bank — it established that between January 2023 and February 2025 they had accumulated at least 158 separate IT failure incidents totalling at least 803 hours of unplanned outage: the equivalent of more than 33 days. And those figures did not even include the Barclays outage described above, or further disruption at various banks on 28 February 2025.
More than a month of downtime, across two years, at the institutions through which essentially the entire population receives its income.
The committee’s chair, Dame Meg Hillier, made the relevant observation: losing access to banking on payday can be terrifying for people living paycheck to paycheck. This is exactly right and worth stating precisely. An outage is an inconvenience if you have a buffer and a crisis if you do not. The same three-day failure produces a mildly annoying weekend for one household and a missed rent payment, a declined card at a supermarket till and a charge for a returned Direct Debit for another.
Compensation followed, unevenly. Barclays expected to pay between £5m and £7.5m for the January incident, bringing its two-year total to as much as £12.5m. The second-highest figure across all nine institutions was £350,000, paid by Bank of Ireland.
That gap should not be read as evidence that eight banks behaved well and one badly. It reflects the fact that there was no consistent standard governing what customers were owed when the system was unavailable. Your entitlement depended on which bank you happened to be with.
Who Pays for Authorised Push Payment Fraud?
The second failure mode is more interesting, because the infrastructure works perfectly and the harm happens anyway.
Authorised push payment fraud is what occurs when somebody deceives you into instructing your own bank to send money to a criminal. You are not hacked. Your password is not stolen. The Faster Payments System does precisely what it was built to do, at high speed and with complete reliability, and the money is gone in seconds — which is the point of the system and, from the fraudster’s perspective, the attraction.
The scale is substantial. UK Finance recorded total fraud losses of £1.28 billion in 2025, with APP fraud rising while unauthorised fraud fell.
For years, the loss lay with the victim. Banks argued, with some legal support, that a customer who authorises a payment has authorised it, and the Supreme Court’s 2023 decision in Philipp v Barclays confirmed that the common law offered victims relatively little. A voluntary industry code produced wildly inconsistent outcomes.
Then, on 7 October 2024, that changed. Under a statutory duty created by the Financial Services and Markets Act 2023, the Payment Systems Regulator imposed a mandatory reimbursement requirement: banks and payment firms must generally reimburse eligible APP fraud victims up to £85,000, normally within five business days, with the cost split equally between the sending and receiving institutions.
Read that last clause again, because it is the whole argument of this section. The regulator did not merely decide that victims should be repaid. It decided that the bank which received the criminal’s money shares the bill — creating, for the first time, a direct financial incentive for institutions to care about who is opening accounts on their platforms.
The first independent evaluation, by Frontier Economics in July 2026, found that in-scope APP scam losses over Faster Payments fell by around 21% — roughly £73 million a year — and that reimbursement of confirmed losses rose from 54% before the policy to 65% after. The predicted moral hazard largely failed to appear: only around 3% of claims in early 2025 were rejected on the grounds that the consumer had not been sufficiently cautious.
It also found something less comfortable. Reimbursement rates across payment service providers ranged from 21% to 94%. Some of that is differences in record-keeping, but some is genuine divergence in how firms apply the exception for insufficiently careful consumers. Whether you get your money back still depends partly on who you bank with.
The sociological point here is larger than the policy. When an abstract system fails, the loss does not evaporate; it lands somewhere. Where it lands is not determined by the technology, or by fault in any ordinary moral sense — the victim was deceived, the bank was not negligent, the network functioned. It is determined by a rule, and rules are made by people, and can be changed. For roughly two decades the rule put the loss on the individual. Since October 2024 it has largely put it on the industry. Nothing about the technology changed.
This is what it looks like when the politics of an infrastructure become visible.
Can Shops Refuse Cash in the UK?
The third failure is the quietest, because it affects people who are not making a transaction at all.
Around three million adults in the UK still rely on cash for everyday payments. The FCA’s own research identifies who they are with uncomfortable clarity: people with low digital capability or poor digital access are around four times as likely to rely on cash, and people in low-income households around three times as likely.
The people most dependent on the older system are, predictably, those with least access to the newer one.
Government and regulators have responded, and the response is genuinely substantial. Under powers granted by the Financial Services and Markets Act 2023, the FCA’s access to cash regime came into force on 18 September 2024, requiring designated banks and building societies to assess gaps in local cash provision, to respond to requests from communities for such assessments, to deliver additional services where significant gaps are found, and to keep branches and ATMs open until replacements are available. Industry committed in September 2024 to opening 350 banking hubs; by March 2026 there were 225 operational. The FCA reports that more than 95% of the population can access cash within the government’s specified distances.
But there is a hole in the middle of this, and the FCA has been explicit about it. Existing law allows retailers to decide whether to accept cash or not, and the FCA cannot require them to do so.
So the regime protects your ability to obtain cash. It does not protect your ability to spend it.
For someone who is digitally excluded, has no smartphone, cannot pass a bank’s identity checks, or manages a tight budget by physically dividing notes into envelopes, this is the difference between a system that works and one that does not. A banking hub in the market square does not help if the café, the car park and the bus have all gone card-only. A Treasury Committee inquiry into cash acceptance has examined precisely this gap, and the government’s response points to the hubs and the access regime — both of which address the supply of cash rather than its acceptance.
The result is a peculiar kind of exclusion, one that nobody legislated and no individual business is responsible for. Each retailer going cashless makes a reasonable decision about handling costs, security and staff time. The aggregate is a payments environment in which the poorest and least connected people hold a form of money that a growing number of places will not take.
Trust, in other words, is not evenly distributed. The abstract systems Giddens described extend their reliability to those already inside them. To be outside is not to be untrusting; it is to be untrusted, or simply unaccommodated.
Susan Leigh Star, who founded the study of infrastructure as a subject in its own right, put the point in eight words: one person’s infrastructure may be another’s barrier. The systems that make life frictionless for most people do so by drawing a boundary, and the boundary has an outside. Star was interested throughout her career in what she called residual categories — the box marked other, the “not elsewhere classified” — and in what happens to the people who land there. Her examples were deliberately mundane: the left-hander in a right-handed world, the chronic disease sufferer in a medical system built for acute illness, the vegetarian in McDonald’s.
Add to that list the person holding notes in a cashless café. The transaction is seamless for everyone the system recognises — that is precisely what makes it good infrastructure — and for the person it does not recognise, it is a refusal that nobody issued. No shopkeeper decided that the digitally excluded should be unable to buy a coffee. Each calculated their own handling costs. The exclusion is an emergent property of a thousand reasonable decisions, which is the hardest kind to argue with, because there is nobody to argue with.
Trust as a Weapon
One further property of trust infrastructure deserves noting, because it operates at the level of states rather than households.
Following Russia’s invasion of Ukraine in 2022, a number of Russian banks were disconnected from SWIFT as part of coordinated sanctions. No money was seized. What was removed was access to the messaging network through which international banking is conducted — the ability to be spoken to, in effect, by the rest of the financial system.
That this works as a sanction tells you what payment networks fundamentally are. They are not pipes carrying value. They are systems of mutual recognition, and exclusion from them is a form of unpersoning at the institutional level.
Predictably, it has accelerated the construction of alternatives, with China expanding UnionPay and promoting its Cross-Border Interbank Payment System for yuan settlement. Whether a fragmented global payments landscape is safer than a concentrated one is a genuinely open question. What is not open is that the concentration was, for a period, a source of real power, and that using it visibly has made others determined not to be subject to it.
Who Bears the Loss
The card terminal beeps and you walk out with your coffee. Behind that sound sits an arrangement of remarkable sophistication and, on the whole, remarkable reliability: billions of transactions a year, a very small proportion of which go wrong.
But reliability is not the interesting property. The interesting property is that any system of this kind must constantly answer a question that has no technical solution: when this fails, who pays?
Britain has answered that question three different ways in three different places, and the answers reveal more about the society than the engineering does. When the bank’s systems go down, the customer largely absorbs it, with compensation varying by an order of magnitude depending on the institution. When a fraudster exploits the system’s speed, the industry now absorbs it, because a regulator decided in 2024 that it should. When somebody cannot participate at all, nobody absorbs it — the loss simply sits with them, distributed across a thousand individually reasonable decisions by shopkeepers.
The systems in this series so far have been allocated by a connection queue, by expected financial return, and by a lease. This one is allocated by eligibility. To use it you must be recognisable to it — hold an account, pass the checks, own the device, fit the pattern that the fraud model expects. Most people do, effortlessly, and never discover the condition exists.
Satellite timing is next in this series, and the sequence is not accidental. Payment networks depend on a shared, precise definition of when a transaction occurred, and that definition comes from somewhere. Having asked who is trusted, we turn to who supplies the clock they are trusted against.
The beep, in the end, is not the sound of money moving. It is the sound of a system saying that it knows who you are, and has decided to proceed.
References
| Claim | Source | URL | Why it supports the claim |
|---|---|---|---|
| At least 158 IT failure incidents and at least 803 hours (33+ days) of unplanned outage across nine banks and building societies, January 2023 – February 2025; figures exclude the Barclays 31 Jan–2 Feb outage | Treasury Committee, UK Parliament | https://committees.parliament.uk/committee/158/treasury-committee/news/205611/more-than-one-months-worth-of-it-failures-at-major-banks-and-building-societies-in-the-last-two-years | Committee’s own published summary of the correspondence it collected |
| Barclays could pay up to £12.5m in total; second-highest payout £350,000 by Bank of Ireland | as above | as above | Stated directly in the committee’s summary |
| Barclays outage 31 Jan–2 Feb 2025: 5% of logins failed; of successful logins attempting payment, 56% failed; cause a software problem in a critical UK mainframe module; expected payout £5m–£7.5m | FinTech Futures, reporting Vim Maru’s letter of 26 February 2025 to the Treasury Committee | https://www.fintechfutures.com/bankingtech/barclays-set-to-pay-out-millions-in-customer-compensation-after-it-outage | Quotes the letter’s figures and stated cause |
| Treasury Committee’s original request to nine named banks; Bank of England view on systemic risk from simultaneous failures | Treasury Committee, UK Parliament | https://committees.parliament.uk/committee/158/treasury-committee/news/205186/bank-outages-committee-demands-answers-from-banks-and-building-societies-following-barclays-it-failure/ | Names the institutions and sets out the committee’s questions |
| Mandatory APP reimbursement from 7 October 2024; up to £85,000; within five business days; cost split equally between sending and receiving PSPs; statutory basis in s.72 FSMA 2023 | Bratby Law, analysis of the PSR reimbursement requirement and its evaluation | https://bratby.law/app-fraud-reimbursement-evaluation/ | Sets out the mechanism, cap, timing, liability split and legal basis |
| Frontier Economics evaluation (July 2026): in-scope APP losses down c.21% (c.£73m/yr); reimbursement up from 54% to 65%; reimbursement rates ranging 21%–94% across PSPs | as above | as above | Reports the evaluation’s findings including the variance |
| c.3% of claims rejected in Q1 2025 for failing the consumer standard of caution | Edgar, Dunn & Company | https://www.edgardunn.com/articles/one-year-on-uks-app-fraud-reimbursement-rules | Cites PSR statistics on rejection rates |
| Total UK fraud losses £1.28bn in 2025; APP fraud losses rose 4% while unauthorised fraud fell 5% | Refundee, reporting UK Finance data | https://www.refundee.com/blog/uk-finance-fraud-report-2025-app-fraud-victims | Reports the UK Finance headline figures |
| £159.2m of APP losses returned to victims in H1 2025 (62% of the total stolen); 88% of in-scope losses reimbursed per PSR | UK Finance | https://www.ukfinance.org.uk/news-and-insight/press-release/over-ps600-million-stolen-fraudsters-in-first-half-2025 | UK Finance’s own press release with the figures |
| Access to cash regime in force 18 September 2024; duties to assess gaps, respond to community requests, deliver additional services and keep facilities open | FCA, PS24/8: Access to cash | https://www.fca.org.uk/publications/policy-statements/ps24-8-access-cash | The FCA’s final policy statement |
| Around 3 million adults rely on cash; the FCA cannot require retailers to accept cash | FCA press release on the final rules | https://www.fca.org.uk/news/press-releases/fca-confirms-plan-protect-access-cash-consumers-and-small-businesses | States both the figure and the limit on the FCA’s powers |
| Cash reliance 4x more likely with low digital capability, 3x more likely in low-income households | Hogan Lovells, summarising the FCA’s research underpinning PS24/8 | https://www.hoganlovells.com/en/publications/uk-access-to-cash-fca-publishes-final-rules | Reports the FCA’s research findings on cash reliance |
| 225 operational banking hubs as of March 2026; commitment to 350; FSMA 2023 powers did not extend to wider banking services | House of Commons Library, Access to cash and banking services (CBP-9453) | https://commonslibrary.parliament.uk/research-briefings/cbp-9453/ | Parliamentary briefing with the hub count and legislative background |
| Government response on cash acceptance, 350 hubs commitment, National Payments Vision | House of Commons, Acceptance of cash: Government Response | https://publications.parliament.uk/pa/cm5901/cmselect/cmtreasy/1128/report.html | The published government response to the committee’s report |
| More than 95% of the UK population can access cash within government-specified distances; 121 hubs opened in the regime’s first year | FCA | https://www.fca.org.uk/news/press-releases/communities-use-rules-successfully-maintain-access-cash | FCA’s first-anniversary statement on the regime |
Note on sourcing. Where a claim rests on correspondence to a select committee, I have cited the committee’s own published summary in preference to press coverage of it; the underlying letters are available from the same page. UK Finance is an industry body and its fraud data is self-reported by members, which is worth bearing in mind — though it remains the only comprehensive series available, and the PSR’s independent evaluation broadly corroborates the direction of travel.
Reading
- Anthony Giddens, The Consequences of Modernity (Polity, 1990) — symbolic tokens, expert systems and trust in abstract systems, pp. 79–100
- Geoffrey C. Bowker and Susan Leigh Star, Sorting Things Out (MIT Press, 1999) — residual categories and the cases that do not fit
- Georg Simmel, The Philosophy of Money (1900) — money as a form of socialised trust
- Viviana Zelizer, The Social Meaning of Money (1994) — on how people mark and separate money in practice
- Bill Maurer, How Would You Like to Pay? (2015) — on the anthropology of payment infrastructure
How Modern Society Works
← Previous: What Is Cloud Computing? · Series hub · Next: How GPS Works →
