Europe is losing roughly €480,000 an hour to payment fraud. I got that figure by dividing the €4.2 billion the European Banking Authority (EBA) and the European Central Bank (ECB) reported across the European Economic Area in their latest joint report by the number of hours in a year; the total covers card fraud, not just scams. The credit transfer numbers are the ones that stay with me. Customers bore about 85% of those losses, and in most cases nobody broke into anything. The scammer talked, and the customer sent the money.
Then came the Financial Action Task Force (FATF), which has just put out a report on scam compounds, and it explains why so little of that money is ever recovered. The watchdog traced stolen funds moving from a victim’s account to an exit from the financial system in three to five minutes. The average victim needs about 20 hours to work out what happened.
Five minutes against 20 hours. I have spent years in fraud detection, and this is no longer a detection problem. Whether funds can be frozen gets decided in the first 24 to 72 hours after a report, and by the time the report comes in, the money is usually long gone. A bank sees the payment. A messaging platform saw the recruitment. A telecom carried the call. Nobody sees it all, which means the only defense that can keep pace with the scammer is one in which those organizations tell each other what they know while there is still time. Right now, the law does not clearly let them.
I took that question to a European Commission roundtable on online scams this month. I came away encouraged, and also with half of my question unanswered. On November 19, the Commission decides whether to start the one process that could answer it.
The Victim Pressed Send
Most of the payment controls Europe built over the past decade exist to stop a stranger from using your account. They work. The EBA and ECB report says as much: strong customer authentication remains effective against the fraud it was designed to prevent. The same report warns that fraud through payer manipulation is on the rise. A scam payment passes every check, because the account holder is the one who approved it.
This leaves the bank as the last organization to see the scam and the only one anyone expects to stop it. The FATF breaks the scam economy into seven stages. Social media and messaging platforms show up in the first four, from setting up the operation and recruiting workers through finding the victim and executing the payment. By the time money moves, the scammer’s work is mostly done, and most of the evidence of it sits with companies that never touch a payment. Controls that live at the point of payment are reading the last page of a story the platforms watched from the start.
Organized Crime With a Payroll
It helps to drop the word “scammer” for a moment and picture a company. The compounds the FATF describes have recruiters, shift rotas and, in some cases, their own laundering desk down the hall. Investigators have found people from at least 80 jurisdictions working in Southeast Asian compounds, and a large share of them did not choose to be there; the estimate for Myanmar alone is 120,000 people forced into online fraud. Proceeds are mostly left as virtual assets, often via an over-the-counter exchange desk operating within the compound.
Nor does any of this take much skill anymore. Anthropic’s September threat intelligence report has a section titled “Sophisticated attacks no longer require sophisticated attackers,” and the argument underneath it is that intent, rather than technical ability, is now what separates a state-backed operator from anyone else with a laptop.
The FATF’s answer is a list of six principles for governments and industry. I recommend anyone in detection use the third one. It asks financial institutions, technology firms, telecoms and public authorities to connect their networks and share what they know, which, read plainly, is a request to copy the opposition. A recruiter in one country, a mule herder in another and an exchange desk in a third already trade information in real time, and not one of them has ever had to run it past a legal department.
Brussels Is Building the Scaffolding
The Commission is moving. An official told the roundtable that it is preparing a voluntary code of conduct on online scams under Article 45 of the Digital Services Act and will take the question of whether to proceed to the European Board for Digital Services on November 19. Signing up would be voluntary, she said, but signatories’ commitments would be checked and audited, and enforcement carries on either way. Proceedings against Meta over deceptive advertising on Facebook and Instagram have been open since April 2024, and on October 21 the Commission opened an investigation into whether Booking.com properly assessed and reduced the risk of fraudulent listings and phishing.
What struck me most, listening to practitioners and officials over the course of the day, was how little patience anyone had left for process metrics. Nobody wanted to hear how many logins a bank had blocked. They wanted to know how much money had stayed in customers’ accounts because of something the bank did, and several people said, more or less unprompted, that this ought to be the published number: euros intercepted before they left the system, and the time it took to act on the alert. I agree with them. I would also point out that no institution can post a good number on that scoreboard by itself, which brings me to the gap I put to the Commission.
Who Can Share What
Europe is, on paper, already building a channel for this. Article 75 of the new Anti-Money Laundering Regulation allows obliged entities to share information through supervised partnerships, and ahead of those provisions taking effect on July 10, 2027, the European Data Protection Board and the new EU Anti-Money Laundering Authority are jointly drafting guidelines on how such partnerships can operate. The authority has also put out a call for up to 12 technology providers that already support AML information-sharing arrangements to help with the drafting, which I was glad to see. Still, read the scope again: obliged entities, partnering with each other. Where does that leave a provider that sits across thousands of obliged entities at once? Nowhere in particular, as far as I can tell.
I should explain why I keep coming back to that, since it is my day job. We process risk data for banks, lenders, payment firms and marketplaces, and it is an ordinary week for us when a single device fingerprint (or mule account, or synthetic identity) shows up at a client in Lisbon, then at one in Warsaw, then a third in Dublin. Suppose Lisbon confirms it as fraud. Warsaw and Dublin could, in principle, be protected within minutes. They usually are not, and the obstacle is one sentence in the EDPB’s guidance on controllers and processors: a provider handling data on a client’s behalf may use it only for that client’s purposes. Take what Lisbon taught you and use it to protect Warsaw, and you have, arguably, made yourself a controller, with everything that entails. I know of few providers willing to carry that risk on an untested legal theory, even when Lisbon and Warsaw are both asking them to. So the company with the widest view of a fraud ends up with the clearest instruction to keep quiet about it.
I put roughly that to the Commission. The official replied that the General Data Protection Regulation (GDPR) is not a blanket ban, that legitimate interest can support fraud-related sharing in the right circumstances, and that a code could specify who shares what and for what purpose. I do not quarrel with any of it. But I asked a follow-up, whether anyone was working with the data protection authorities on the position of intermediaries specifically, and the answer was no. Until that changes, a provider’s lawyers have nothing to cite but a sympathetic remark at a roundtable.
My second question didn’t get an answer so I will put it here instead. Like any processor, we receive data subject access requests, and a proportion of them come from people one of our customers has just flagged for fraud. Nothing in the GDPR, and nothing in the guidance on access rights, says how a provider is supposed to respond to one of those without handing over the list of signals that caught the person asking. Unless someone says otherwise, the legally safe approach is to disclose. Which means the law, as it currently reads, obliges us to run a tutorial for fraud rings on request.
What a Code Should Settle
If the Commission goes ahead on November 19, the code is where this gets fixed. I do not think the fix is complicated, but it has to be written down somewhere. Sharing a confirmed fraud signal through a processor needs to be named as a legitimate purpose, and someone has to define “confirmed,” because that one word will carry most of the weight. Give the signal a short retention limit, and an audit trail a regulator can pull whenever it likes; I would happily accept both. Get the data protection authorities into the drafting rather than consulting them afterward, so that “we are waiting for clarity” stops being a respectable reason to sit on something that could have saved a customer. Somebody has to say, in writing, how the GDPR’s existing power to restrict access rights in the interest of preventing crime applies to fraud signals, so that a flagged fraudster’s access request is not answered with the evidence against him. And the code should ask signatories to publish the yardstick the roundtable kept asking for: euros intercepted and time to act, because once that is the number people watch, holding a signal back starts to look like what it is.
I know what the objection is, because I have heard it in every version of this conversation: privacy. So let me describe what I am asking for in full. A confirmed fraud signal. Kept for a short, fixed period. Logged and open to inspection by a regulator at any time. Compare that with the behavioral data the recruitment platforms already hold on every one of their users, and it is a modest ask. Data about fraud will move regardless of whether Brussels acts. The only open question is whether it moves under rules.
The criminals already have the connected network the FATF is describing. We can see much of what they do, spread across thousands of client accounts, and a legal grey area keeps each piece where it landed. Brussels decides on November 19 whether to start closing that gap. In the meantime, the clock runs at about €480,000 an hour, and somewhere in Europe, another person is pressing send.
