AI-enabled banking fraud to treble to 2.2bn transactions by 2031

Juniper Research warns agentic AI is industrialising social engineering, pushing fraudulent banking transactions past 2.2 billion within five years.

By Disrupts Editorial Team Mon, 28 September 2026 AI level 2 of 5

AI level 2 of 5: drafted by our AI editorial assistant from the company's announcement; fact-checked, edited and signed off by Mark Walker, Editorial Director. What the levels mean

A brightly lit data center aisle features parallel rows of black server racks displaying glowing green and blue indicator lights, extending into the distance under overhead fluorescent lighting.

Fraudulent transactions in digital banking and money transfer are on course to exceed 2.2 billion globally by 2031, up from 773.7 million in 2025, according to new research from Juniper Research. The Hampshire-based market intelligence firm says the near-180% rise is being driven not by attacks on banking infrastructure but by a structural shift towards targeting end-users directly, with artificial intelligence dramatically reducing the cost and complexity of executing personalised social engineering at scale.

The shift matters because it upends the traditional fraud-prevention calculus. Banks have historically focused controls on the transaction layer, monitoring for anomalous payment behaviour after a customer has already been manipulated into authorising a transfer. Juniper's findings suggest that model is no longer adequate when the attack itself is cheap, adaptive, and targeted at the authenticated customer journey rather than the ledger beneath it.

Agentic AI changes the threat geometry

The report draws a distinction between generative AI, which is already enabling more convincing and varied scam typologies, and the emerging layer of agentic AI, which allows fraudsters to coordinate multi-stage attacks and adjust tactics dynamically as victims or fraud controls respond. Where a phishing campaign once required a human operator to monitor and adapt, agentic systems can do so autonomously and continuously, making fraud operations more persistent and harder to fingerprint.

Report author Shane O'Sullivan put the challenge in operational terms: "Coupled with a reduced cost to commit fraud, AI enables fraudsters to create and adapt attacks faster, while instant payments reduces the window that banks have to identify suspicious behaviour. Banks must therefore combine behavioural, identity and payment intelligence prior to authorisation, rather than relying on controls after a payment has been initiated."

The instant-payments dynamic is significant. Real-time settlement, now mandated or strongly encouraged across major markets including the EU and the UK, compresses the fraud-detection window to milliseconds. Banks that designed their controls for batch-settlement environments are now operating in a regime where the fraud is effectively complete before a human analyst can review a flag.

Cross-sector and capital implications

The convergence here is threefold. First, the same agentic AI tooling being explored by enterprise software vendors, defence contractors, and logistics operators to automate complex decision-making is simultaneously available to criminal networks at commodity pricing. The dual-use nature of frontier AI is no longer a theoretical policy concern; Juniper's forecast quantifies it in transaction volumes.

Second, the fraud-prevention technology market itself becomes a significant capital destination. Banks under regulatory pressure to demonstrate pre-authorisation controls will need to procure or build continuous risk-profiling systems that integrate behavioural biometrics, identity signals, and payment-pattern analysis in real time. That creates procurement opportunity across cybersecurity, identity verification, and AI-infrastructure vendors, sectors where investment activity has already been elevated by enterprise AI adoption cycles.

Third, the macroeconomic exposure is material for insurers and central banks alike. As fraud losses migrate from infrastructure compromises (which banks typically absorb) to authorised push-payment scams (where liability frameworks vary and regulatory reimbursement obligations are growing, particularly in the UK), the loss burden is being redistributed across the financial system in ways that affect capital adequacy models and insurance underwriting. Regulators in the UK and EU are already revisiting who bears the cost of APP fraud; a near-trebling of transaction volumes by 2031 will accelerate that political pressure considerably.

The Juniper report covers 44,700 datapoints across a five-year forecast horizon and includes a vendor leaderboard for the fraud-prevention market. The full study and a free extract are available via Juniper Research's website.