Skip to content
AI Gives Fraudsters “Advanced Intrusion” Capabilities, Visa Threat Report Finds

AI Gives Fraudsters “Advanced Intrusion” Capabilities, Visa Threat Report Finds

Fintechnews.Sg • November 19, 2026

AI is helping criminals carry out fraud attacks faster, at a greater scale, and with far greater precision, according to Visa’s latest Fall 2026 Biannual Threat report.

Threat actors are putting AI and agentic systems to work at several stages of an attack, from scouting targets automatically to refining malicious code and testing payment controls.

Visa’s Payment Ecosystem Risk and Control (Visa PERC) shares that while these fraud techniques aren’t new, with AI, threat actors now have “advanced intrusion” capabilities, regardless of their available resources, expertise, or geographic location.

AI Helps Fraudsters Find Weak Links in Payment Supply Chains

Over the months of Jan- June 2026, Visa PERC has seen fraud attacks become more sophisticated across the global payments ecosystem, driven largely by AI. Threat actors are using AI tools and models to research victims and spot weaknesses they can exploit as a way in.

The threat actors’ main targets are third-party providers and other players across the supply chain that serve many merchants downstream or process payments for multiple businesses. By breaching one of these, criminals can compromise payment account data and use it to commit fraud.

Incidents Involving Third-Party Agents Rise 62%

Threat actors are getting better at exploiting vulnerable third-party providers, using them to reach the largest pool of payment account data for the least effort. Incidents involving third-party agents rose 62% compared with the six months.

The report also flagged systemic risks on e-commerce platforms that onboarded merchants without adequate vetting, monitoring or enforcement controls in place. These gaps let threat actors bring fake or collusive merchants into the payments ecosystem, which puts acquirers, processors, and the wider industry at risk.

Additionally, Visa PERC investigated and acted on numerous large data breaches, analysing millions of recovered accounts. Recovered Primary Account Numbers (PANs) from data breaches rose more than 1,200% in January to June 2026 compared to the six months.

Recovered payment accounts were the main driver of compromised PANs in this period, compared to other methods like merchant compromise and network intrusion.

AI Aids Fraudsters in Targeting More Victims Across Borders

Visa PERC highlighted that fraud attacks keep growing in scale, as threat actors were able to hit victims more frequently in several geographic regions for the same fraud campaign.

AI agents and models have greatly strengthened their hand, especially in targeting cardholders and organisations in ransomware attacks and scams, regardless of the language the victims speak.

Aside from automating reconnaissance and finding vulnerabilities, AI can help threat actors target specific groups like the elderly with cryptocurrency investment and romance scams.

It can also identify senior IT staff at target organisations who may have access to the Secure Shell (SSH) accounts used for user authentication. Attackers can then design spear-phishing campaigns, highly targeted social engineering attacks to steal their credentials and carry out double extortion ransomware attacks.

Visa Identifies US$1.4 Billion in Fraud as Scam Networks Expand

Scam-enabled commerce kept spreading across the payments ecosystem over the past six months, according to Visa Scam Disruption. Large-scale e-commerce schemes, subscription traps, merchant network abuse, unauthorised recurring billing, and social media deception are increasingly driving fraud.

Recurring billing and deceptive e-commerce persisted as dominant threat vectors. Visa Scam Disruption also saw growing activity in investment scams, tech-support fraud, transaction laundering, and collusive merchant networks.

Its investigation uncovered interconnected infrastructures spanning multiple regions and acquisition channels. This points to scam operations that run on an industrial scale while also being able to change tactics with agility.

In response, Visa Scam Disruption stepped up its disruption work via acquirer engagement, security alerts, intelligence sharing, government outreach, and enhanced analytics activities. These efforts helped it identify more than US$1.4 billion in fraud, up 40% from the six-month period.

From Deepfakes to Phishing, AI Makes Scams Feel Personal

Visa observed that fraudsters are using deepfakes to get past authentication checks. While this is an increasingly known statement, it also reveals that AI agents and AI tools are making it easier to run phishing campaigns at scale while tailoring messages with details specific to an individual or an entire organisation.

Numerous industrialised scam compounds have integrated generative AI, LLMs and AI agents into their operations. That allows them to approach more potential victims across multiple markets.

Victim can happen via an unsolicited call, messaging apps, or social media. Fraudsters use AI-generated voices, scripted dialogue, and language translation tools to pose as financial professionals or someone a victim knows.

Payment Fraud is Becoming More Precise

Visa PERC observed a marked increase in advanced intrusions, aimed at specific vulnerabilities or misconfigured security controls. The trends span card-present, card-not-present (CNP), and tokenised channels, with the shift becoming especially precise and targeted in card enumeration, CNP fraud, and abuse of token provisioning.

This possibly hints at the wider availability of AI tools and other advanced technologies. Fraudsters could use AI tools to test and probe payment ecosystem players and learn how authorisation decisions are made, helping them identify how a fraudulent transaction is likely to be approved.

When Frontier AI Models Overstep Their Test Environments

Visa PERC cited public reporting and cybersecurity research describing incidents where frontier AI models, which are systems of advanced AI agents built for proactively performing cybersecurity evaluations, operated beyond the limits of their isolated test or training environments.

The models were still in testing and had been trained to help researchers identify security weaknesses within those environments.

Yet in one reported incident, a model went beyond its protocols, exploited a previously unknown zero-day vulnerability, and gained unauthorised access to an organisation’s datasets. It then exfiltrated the compromised credentials.

In doing so, the incidents show how frontier AI models can act well beyond their instructions while being adept at detecting vulnerabilities organisations don’t know exist. In the hands of threat actors, those same capabilities could be turned against payment firms.

In its assessment, Visa PERC shared that threat actors could use teams of frontier AI agents to replace tens or hundreds of human operators. As these agents can find and exploit unknown vulnerabilities, they could pose a major threat to payments worldwide.

Frontier AI models could potentially help threat actors to pinpoint the most minute gap to abuse within a victim’s network or processes. They could then break in or monetise stolen PAN data.

Paul Fabara, the Chief Risk and Clients Officer at Visa , shared,

“At Visa, we’re fighting bad AI with good AI, combining technology, network intelligence and human expertise to identify emerging threats earlier and help stop fraud before it scales.”

“At Visa, we’re fighting bad AI with good AI, combining technology, network intelligence and human expertise to identify emerging threats earlier and help stop fraud before it scales.”

Visa’s response is to fight AI with AI, using frontier AI models that simulate fraud scenarios that threat actors could uncover through probing, and shifting fraud controls further upstream in the authorisation process.

Featured image edited by Fintech News Singapore based on an image by designer vibe on Magnific

Extracted Entities