Fighting payment fraud with AI

For business leaders, payment fraud is not just a nuisance: it’s a strategic risk. Traditional defences are being stretched by criminals who are using generative artificial intelligence (gen AI) to scale their attacks. 

Yet AI is also rapidly becoming one of the most powerful tools in the fight against fraud. What was once a niche technology is now critical for protecting revenues, improving customer experience and preserving trust.

The state of payment fraud

A recent report by global payments processing platform Stripe, 2025 State of AI and Fraud, paints a stark picture. Fraudsters are no longer working in opportunistic ways: they are acting with industrial efficiency. They use gen AI to create fake identities and build entirely artificial websites, constructing highly convincing fronts to solicit payments. According to Stripe, one in three business leaders say that AI is making fake account creation and merchant fraud worse.

Meanwhile, more conventional fraud methods, such as brute-force attacks and credential stuffing, continue, but are now being automated and scaled. This “card-not-present” fraud is amplified by machine-speed attacks, which legacy systems were never built to counter. The consequences go beyond the immediate financial hit. Fraud also inflicts second-order damage: chargebacks, operational costs associated with manual reviews, reputational risk and, perhaps most insidious of all, false declines.

False declines add extra damage

In response to heightened fraud risk, many businesses have understandably ratcheted up their security processes. Often, they will rely on rules-based systems to weed out fraudulent payments: “if AVS fails, decline”; “if CVC mismatch, block”; and so on. This approach may stem the flow of fraud, but it does so at a cost. Rules-based systems tend to err on the side of caution because of their rigid logic: with static rules, there is no nuance, no real-time adaptability to fraudsters’ evolving tactics.

The result, in many cases, is a rise in false positives, with legitimate transactions being flagged and declined. These false declines don’t make headlines in the way major cyber-attacks do. But they can inflict even greater long-term damage. When genuine, honest customers are blocked, that’s not only lost revenue but also a very negative experience that can destroy loyalty and drive customers elsewhere.

Why traditional fraud prevention is failing

Traditional fraud solutions, such as rule engines and blacklists, worked well in a slower, less adversarial world. But fraud operations today, especially those powered by AI, pose a particular challenge to rule-based defences. Rules engines cannot easily detect sophisticated fake identities created by gen AI, nor can they dynamically adapt when fraudsters change strategy.

With fraud evolving so rapidly, static processes cannot keep pace with constantly changing patterns of abuse. And manual review teams are overwhelmed: the velocity and volume of modern e-commerce make the manual checking of digital payments costly, slow and prone to error.