Next-Generation AI Means Next-Generation Fraud: Are We Ready?
Dan Muller, CEO of payment technology company Aeropay, believes that AI will not automatically solve fraud problems but may instead be maliciously exploited, giving rise to new attack methods such as deepfake identities and voice cloning. He calls on the industry to shift focus from pursuing advanced models to building trust-based judgment and defensive infrastructure to address next-generation fraud risks.
Editor's note:Dan Muller is the CEO of Aeropay, a payments fintech company headquartered in Chicago.
Currently, the discussion around AI and payments is booming. Claims about fraud detection becoming smarter, checkout processes fading into the background, and "AI-driven" narratives are flooding nearly every feed.
Some of these claims are true, but AI will not save us from fraud. It cannot stop malicious actors from impersonating others, and it cannot protect your customers when deepfakes are convincing enough to bypass security verification.
We can already see the beginnings of the future: synthetic identities, cloned voices. These are no longer hypothetical scenarios, but clear evidence that AI is being rapidly used to break through the systems designed to protect us.
Can deepfake AI pass "Know Your Customer" (KYC) verification? What happens when a model can mimic a face, blink in front of a camera, and read out a driver's license number aloud? What if AI tools can generate synthetic merchants with seemingly legitimate websites and transaction records?
There is also the human factor. Imagine a cloned voice calling a customer service hotline. It sounds identical to the account holder, complete with background noise and tone. How confident are we that a customer service representative can see through this deception before processing the request?
Yet, the current narrative continues to point to more innovative models as the solution, rather than confronting the obvious problem.
AI will not choose your partners for you. It will not build your fraud control system. It also does not care whether you end up trusting the wrong system. And this is the part that receives insufficient attention.
Deepfake-level impersonation, or agentic AI, is no longer a niche issue. It is approaching all systems built on trust.
Imagine a fake merchant successfully onboarding before human review, or an AI model trained on customer service interaction data that learns to steer conversations toward approval outcomes.
The same innovation that makes payments more efficient has also given rise to more sophisticated fraud. So, while everyone talks about how AI will make transactions faster, safer, or more personalized, the real question should be: when fraud methods also become smarter, who ensures these promises still hold?
The next phase of innovation will be about discernment, not necessarily about having the most advanced model. It is about knowing who to trust and which partners are capable of withstanding the risks brought by new technologies.
AI will undoubtedly continue to shape the payments industry. But the companies that will succeed are those that address the more complex problems and impacts AI brings, rather than merely using it as a marketing gimmick.
Who builds the systems you rely on? How do they respond when failures occur? Do they treat fraud as a public relations issue or a product issue? These answers matter more than another round of "AI-driven" headlines.
From my (human) perspective, AI can make payments smarter, but what truly keeps payments running is partnerships and the right infrastructure.