(8 Minute Read)

9 July 2025 / Written by guest contributor Katy Worobec

<aside> 🗞️ TL;DR. Katy dissects AVIEL’s A.I. assisted analysis on more than 4000 scam cases published by the Financial Ombudsman Service (FOS), and reflects on the possible causes of (and solutions to) ‘The Revolut Crypto Surge’ of Summer 2024: a massive jump in FOS cases relating to crypto-related scams. She ponders how PSPs can be crypto-friendly whilst maintaining adequate fincrime controls.

</aside>

If you have nothing better to do on a long train journey, delving into the world of The Financial Ombudsman Service decisions always makes interesting reading (yes, really!).  In my view, it’s required reading if you are a financial institution that is looking to up your game in terms of customer service relating to fraud and scams. But with thousands of decisions on the website, it's difficult to discern any trends from the mass of documents. That, of course, is where the brave new world of AI and specifically Large Language Models (LLMs) come into their own. The talented data team at AVIEL downloaded all the scam and fraud related FOS decisions from the beginning of 2024 through to April 2025 - over 4,000 cases - and used LLMs to have a closer look at the trends.[*1]

First, a note of caution: most complaints come to FOS after the customer has exhausted all options at their bank, and cases will take some months to be examined and a decision made. Some decisions are also appeals from older cases. This time lag means any patterns will most often reflect the state of fraud and scams a year or so earlier, perhaps, than the dates on which the details were published by the FOS.

That said, the treasure trove of Ombudsman reasoning throws up some interesting insights which reflect how the fraudsters are operating and what, in turn, FOS believes the FI can be expected to do to protect customers.

15 Months of Scam Decisions: The Headlines

What really caught my interest was the volume of decisions relating to crypto investment scams throughout 2024, and even more intriguing, the huge surge in cases involving Revolut in the latter half of 2024.

Figure 1: A timeseries showing that starting in August 2024 the FOS started publishing a 5-fold or more increase cases against Revolut when compared to the 4 other PSPs most often subject to FOS complaints about fraud.

Further analysis of the dataset more broadly began to build a picture of some possible, largely independent, correlations:

• Investment scams and crypto currency investment scams together formed 40% of the scam types in the sample. It’s hardly surprising that more investment cases end up reaching FOS: these type of scams are less likely to be reimbursed in the first instance by individual FIs (around 50% in 2024, according to UK Finance [*2]. However, investment/cryptocurrency scams were only upheld (in favour of the victim) by FOS in 34% and 27% of the cases respectively, indicating that perhaps the FOS isn’t proving the silver bullet that the large volumes of consumers making complaints to them would wish.
• Job and Employment scams often have striking similarities with crypto investment scams when underlying victim payments are considered – a whopping 87% mention cryptocurrency and the uphold (in favour of victim) rate is within 1% of that of crypto investment scams at 34%.
• If you are a customer from a challenger bank you are more likely to have your case upheld - nearly 40% of cases from challenger banks were upheld, as opposed to just over 25% from traditional banks.
• Uphold rates were higher (over 40%) where the reported source of scam engagement is from an email or telephone call. Lower uphold rates were seen where the scam came from personal contacts and - perhaps surprisingly at first glance - dating platforms. This is more complex to unpick. Certainly a key factor for FOS is what opportunity the FI has to detect and prevent a scam. Existing tools PSPs have to hand are more likely to detect activity related to someone being manipulated by a scammer in real-time (like in telephone-based impersonation cases), but are less likely to detect someone who may be making low-level payments over a period of time to a ‘known’ payee, a typical pattern in romance scams.
• More controversially, scams originating from social media platforms and digital advertising have relatively high FOS uphold rates (37 – 40%). The UK Finance figures tell us that over 70% of fraud scam cases originate from online platforms, so the FOS figures further underline the need for online platforms to do more to detect and prevent scams up-stream and for FIs to bolster defences against scams leveraging online platforms as a start point.