The global banking industry is heading into a $66.2 billion fraud-defence bill as artificial intelligence gives financial criminals a faster, cheaper and more sophisticated way to attack customers.
Spending by banks and other financial institutions on fraud-prevention solutions is forecast to rise from $36.2 billion in 2026 to $66.2 billion by 2031, an increase of 82.9 per cent, according to new research by Juniper Research.
But the investment will be chasing a problem that is expanding even faster.
Juniper forecasts that fraudulent transactions across digital banking and money-transfer services will rise from 773.7 million in 2025 to more than 2.2 billion by 2031 — an increase of more than 180 per cent.
The numbers point to an emerging arms race in financial services.
On one side are banks, fintechs, payment companies and regulators spending billions to detect suspicious activity.
On the other are criminals using increasingly accessible artificial-intelligence tools to automate scams, personalise attacks and manipulate customers into authorising transactions themselves.
For financial institutions, the most dangerous fraud may increasingly be the transaction that looks completely legitimate.
The criminal no longer needs to break into the bank
The old image of bank fraud is increasingly outdated. The conventional attack involved stealing credentials, compromising systems or exploiting technical vulnerabilities.
The new attack can be much simpler. A criminal persuades a customer that an urgent payment is necessary.
The customer opens the banking application; the customer authenticates; the customer transfers the money.
The bank’s systems see a genuine customer using a genuine device to make an authorised payment.
The criminal has not necessarily defeated the bank’s security system. The criminal has used the customer to defeat it.
This is the significance of Juniper’s warning that artificial intelligence will make social-engineering attacks increasingly sophisticated.
Generative AI can help criminals create convincing, personalised communications at scale, while emerging agentic AI could automate multiple stages of an attack and dynamically alter tactics in response to a victim or a financial institution. The economics are compelling for criminals.
An attack that previously required research, language skills, human operators and time can increasingly be automated.
AI effectively turns fraud from a collection of individual scams into something closer to an industrial process.
That is why the growth in defensive spending matters.
$66bn is the price of fighting back
The $66.2 billion forecast is more than a technology-market statistic.
It is a measure of how expensive trust is becoming in digital finance.
Banks have spent years removing friction from payments.
Customers expect transfers to happen instantly. They expect mobile applications to recognise their devices. They expect authentication to be simple. Businesses expect money to move at any hour.
Those improvements have transformed banking.
They have also created a highly attractive environment for criminals.
The faster legitimate money moves, the faster stolen money can move.
The more customers trust digital channels, the more valuable it becomes for criminals to impersonate those channels.
And the more transactions become automated, the harder it becomes for conventional human review to keep up.
Juniper’s forecast suggests financial institutions understand the scale of the problem.
The industry will spend almost twice as much on fraud prevention in 2031 as it does in 2026.
The question is whether the spending can keep pace with the technology available to the attackers.
Nigeria is already fighting the battle
Nigeria provides one of the clearest examples of the contradiction at the heart of the global fraud story.
The country’s digital-payment fraud losses fell by 51 percent in 2025 to ₦25.85 billion, from ₦52.26 billion in 2024, according to the Nigeria Inter-Bank Settlement System.
Reported fraud incidents also fell to 67,518 in 2025, from 123,918 in 2021. This represents substantial improvement.
It demonstrates that stronger controls, better monitoring and collaboration among financial institutions can reduce fraud losses even as digital payments continue to expand.
But the threat itself is changing.
NIBSS identifies social engineering as the most prevalent fraud technique, alongside SIM-swap attacks, phishing, account compromise and insider abuse. It also reported that industry controls prevented about ₦20 billion in potential losses in 2025.
Nigeria, in other words, may be getting better at stopping yesterday’s fraud while confronting a new generation of attacks.
That distinction is crucial.
A decline in reported losses does not mean the underlying threat is disappearing.
It means the defenders may currently be winning some battles.
The attackers are changing the battlefield.
Â
The Nigerian opportunity — and vulnerability
Nigeria’s rapidly expanding digital-payment economy makes the issue particularly important.
The Central Bank of Nigeria (CBN) says close to 11 billion transactions were processed in Nigeria in 2024, compared with five billion in 2022, while real-time payment channels accounted for more than a quarter of electronic transactions.
That growth is economically significant.
It allows businesses to receive payments faster, reduces dependence on cash and gives fintechs room to build new financial products.
But it also creates an enormous target.
Every additional digital account, payment channel and connected customer creates another potential route for fraud.
Nigeria’s payment ecosystem is therefore entering a more difficult phase: scale is no longer enough.
The system must scale securely.
The real battlefield is behavioural data
This is where the next generation of fraud technology will matter.
Traditional fraud systems ask questions such as:
- Is the transaction unusually large?
- Is the device unfamiliar?
- Is the location suspicious?
Those questions remain useful.
But sophisticated fraud increasingly requires a different question:
Does this transaction make sense for this customer, given everything else we know?
A customer who normally sends small domestic transfers suddenly makes a large payment to a new beneficiary.
A device changes.
The account’s behaviour changes.
The customer’s interaction pattern changes.
A new beneficiary is added shortly before a large transfer.
Individually, none of these events necessarily proves fraud.
Together, they may reveal a pattern.
Juniper is therefore advocating greater integration of behavioural, identity and payment intelligence before transactions are authorised, rather than relying primarily on controls after money has begun moving.
That could fundamentally alter how banks think about security.
Fraud detection becomes less like checking a transaction and more like continuously assessing risk.
But there is a price for being too safe
The danger for banks is obvious.
The easiest way to reduce fraud is often to introduce more friction.
Require another authentication; delay a payment; block an unfamiliar beneficiary; ask more questions; send the customer to a branch.
But every additional barrier weakens the proposition that made digital banking attractive in the first place.
For fintechs, the problem is even more acute.
Many fintech business models are built around speed, convenience and low friction. A system that stops too many transactions may protect the balance sheet while damaging customer acquisition and retention.
The industry’s challenge is therefore not simply to build stronger walls.
The $66bn market is only the beginning
The financial implications extend beyond banks buying fraud software.
Juniper’s research covers a growing ecosystem of technology providers developing tools for identity intelligence, transaction monitoring, behavioural analytics and artificial-intelligence-based fraud detection.
That creates a major opportunity for technology companies. As fraud becomes more sophisticated, fraud prevention becomes a permanent technology investment rather than an occasional compliance project.
Banks will need systems capable of learning. They will need richer identity data. They will need better real-time monitoring. They will need faster intelligence-sharing across institutions. And increasingly, they will need to understand the customer’s behaviour rather than simply authenticate the customer’s credentials.
The industry is consequently moving towards a model in which trust itself becomes a data product.
The striking lesson from Juniper’s numbers is that the financial industry is not facing a single fraud problem.
It is facing an accelerating technology contest. AI gives criminals cheaper tools for attacking customers. The same technology gives banks better tools for detecting those attacks.
Nigeria’s experience captures the paradox. Fraud losses fell sharply in 2025, showing that better controls can work. But social engineering is becoming more prominent, while the country’s enormous real-time payment infrastructure means the consequences of a successful attack can be immediate.
The next phase of the battle will therefore be fought before the payment is made.
Banks will have to determine not simply whether a customer is genuine, but whether the customer’s decision is genuine.
That is a much harder problem and it explains why financial institutions are preparing to spend $66.2 billion globally on fraud prevention by 2031.
The money is not merely being spent to protect bank accounts. It is being spent to preserve confidence in the digital financial system itself.
For Nigerian banks and fintechs, that may be the most important lesson of all.
The future of banking security will not be determined solely by who has the strongest firewall or the most sophisticated authentication.
It will be determined by who can recognise a fraudulent decision before the customer makes it.
The fraudster is using AI to make deception cheaper. The bank’s response will be to make detection smarter.
The institution that wins that race will not simply lose less money, it may win the customer’s trust and ultimately, the customer’s business.
