trending
Artificial intelligence is rapidly changing the economics of online fraud, allowing criminals to automate attacks, exploit stolen identities and test compromised payment credentials at a scale that would previously have required sophisticated criminal networks.
AI-Powered fraud surges
Signifyd’s State of Fraud Report 2026 describes the development as a tipping point for ecommerce, with AI effectively “industrialising” fraud. Overall ecommerce fraud pressure increased 33% year-on-year during the first four months of 2026, while first-party fraud and abuse rose 9%.
The concern for merchants and payment providers is not simply that fraud is increasing. It is becoming faster, cheaper and considerably easier to scale.
Card Testing Accelerates 175%
One of the clearest examples is card testing, where criminals make transactions using stolen payment credentials to establish whether cards remain active before attempting larger purchases.
Signifyd recorded a 175% year-on-year increase in card-testing attacks during the first four months of 2026. AI-powered automation means credentials can potentially be tested continuously and at enormous scale, significantly shortening the window available to issuers, merchants and fraud prevention systems to identify an attack.
There is no shortage of compromised credentials entering the criminal ecosystem. Nearly 14.5 million stolen credit cards were listed on illicit marketplaces during 2024, representing a 20% annual increase, according to data cited by Signifyd.
Account Takeover Enters a New Phase
Account takeover (ATO) is undergoing a similar transformation. Signifyd recorded a 78% year-on-year increase in account takeover attempts between January and April 2026.
Fraudsters can use automated phishing, credential stuffing and other techniques to gain control of established customer accounts. Once inside, they benefit from legitimate transaction histories and stored payment credentials, potentially making fraudulent activity harder to distinguish from genuine customer behaviour.
The report also highlights the growing importance of social engineering. Criminals are increasingly using AI to improve phishing emails, text messages and fraudulent calls, while deepfakes can provide convincing voices and faces for impersonation attacks and synthetic identities.
Fraud Prevention Must Match Machine Speed
The wider financial impact is becoming difficult to ignore. Signifyd cites Interpol estimates suggesting AI-assisted fraud schemes can be 4.5 times more profitable than conventional approaches. Meanwhile, AI-generated scams alone cost consumers almost $900 million during 2025, according to FBI figures referenced in the report.
For the payments industry, this creates an uncomfortable asymmetry. Criminals can increasingly automate attacks around the clock while simultaneously using AI to make those attacks more convincing.
Fraud prevention therefore has to evolve beyond static rules and retrospective detection. Real-time behavioural analysis, identity intelligence and automated risk decisions will become increasingly important as merchants and payment providers attempt to distinguish legitimate customers from machines capable of imitating them.
The emerging fraud battle is consequently becoming one of machine versus machine — and speed is becoming as important as accuracy.
Comments