USA
Report fraud to your bank first, then at ReportFraud.ftc.gov (opens in new tab).
Real World: Money
How Banks And Governments Use AI To Stop Fraud
Banks and governments use computer systems, including machine learning, to spot payments that look wrong and stop them. As examples, the US Treasury says these tools helped prevent and recover over $4 billion in one year, and UK banks stopped £1.45 billion of unauthorised fraud in 2024.
Last checked 16 October 2026. Every figure on this page links to its source below.
How Does AI Spot Fraud?
Every card payment and bank transfer leaves a trail of details: the amount, the time, the place, the device and who is being paid. Fraud systems compare each new payment with your usual pattern and with known scams. Machine learning, a type of AI, gets better at this by learning from millions of past payments, including ones that turned out to be fraud.
When a payment looks unusual, the system can block it, hold it, or ask you to confirm it. That is why your bank sometimes texts you or shows a warning in its app.
What Has The US Treasury Done With AI?
The US Department of the Treasury pays out huge numbers of government payments every year, from tax refunds to benefits. In October 2024 it said its fraud prevention work had prevented and recovered over $4 billion (about £3 billion) in fraud and improper payments in the year to September 2024. The year before, the figure was $652.7 million.
Machine learning AI, used to find fraud involving Treasury cheques faster, accounted for $1 billion of that, in money recovered.
US Treasury fraud prevention, year to September 2024
Flagging high-risk payments$2.5 billion
Finding cheque fraud with machine learning AI$1 billion
Wider risk-based screening$500 million
Changes to payment timing$180 million
How Much Fraud Do UK Banks Stop?
UK Finance, the banking trade body, publishes yearly fraud figures. Its 2025 report says criminals stole £1.17 billion (about US$1.6 billion) through payment fraud in 2024. Banks prevented £1.45 billion of unauthorised fraud, where money is taken without the customer's permission. UK Finance says that equals 67p in every £1 of attempted unauthorised fraud.
The report also found that 70% of authorised push payment cases, where people are tricked into sending money themselves, started online. Another 16% started through phone calls and texts.
Why Does Fraud Still Get Through?
Systems are good at spotting a stranger using your card in another country. They find it much harder when you make the payment yourself because a scammer has talked you into it. To the bank, it can look like a normal payment from your own phone. That is why scam calls and fake messages are still so common.
What It Means For You
- If your bank blocks a payment or asks you to confirm one, check it in your banking app or call the number on the back of your card. Do not use a number from the text or call.
- A real bank will never ask you to move money to a "safe account".
- If a payment warning appears, stop and read it. It may be the system doing its job.
UK
Call your bank on 159, which connects you safely to most major UK banks. You can also report it to Report Fraud.
Common Questions
Is my bank using AI on my account?
Most large banks use automated systems to check payments for fraud. UK Finance says banks prevented £1.45 billion of unauthorised fraud in 2024 using advanced security systems. Ask your bank if you want details.
Why was my genuine payment blocked?
Fraud systems sometimes stop real payments that look unusual, such as a large first payment to someone new. Confirm it through your bank's app or the number on your card.
Did AI really save the US government $4 billion?
Not all of it. The Treasury said its wider fraud work prevented and recovered over $4 billion in the year to September 2024. Machine learning AI was credited with $1 billion of that.
What is 159?
159 is a UK phone number that connects you to your own bank if you think a call or message is a scam.
New to all this? Our free AI Starter Guide walks you through the basics, and you can print it.
Read Next
Sources
Last checked 16 October 2026.
Updated 19 October 2026