By Douglas Hall, Founder & CEO, EQUE
The payments industry is entering a new chapter. AI-driven commerce – where autonomous AI agents shop on behalf of consumers – is moving from theory to reality fast. And, AI fraudsters are moving even faster.
Google has unveiled its Agent Payments Protocol (AP2), designed as “a common language for secure, compliant transactions between agents and merchants.” Visa and Mastercard are getting their networks AI-ready. Worldpay is retooling its systems to support what it calls “agentic” AI shoppers.
The promise is clear: consumers gain seamless purchasing, merchants reach customers in new ways, and payment processors can streamline authorization and intent verification. However, these same innovations carry significant fraud risk if we fail to build controls at the point of purchase.
Innovations Driving the AI Agent Shopping Shift
Here are four AI agentic payment perspectives:
- Google’s AP2 protocol. Stavan Parikh, VP and GM of Payments at Google, and Rao Surapaneni, VP at Google Cloud, describe AP2 as capturing “tamper-proof, cryptographically signed mandates” that detail purchase intent, cart items, and price limits. Partners include Adyen, Mastercard, PayPal, and Worldpay. PayPal’s global head of AI, Prakhar Mehrotra, called AP2 “a strong, adaptable foundation to collectively address the challenges accompanying agentic AI commerce.”
- Visa Intelligent Commerce. Visa opens its payment network to fintech and payment developers and engineers.
- Mastercard tools up agentic commerce. Mastercard’s new tools advance AI-powered payments.
- Worldpay readiness. Cindy Turner, Worldpay’s Chief Product Officer, noted that in fully AI-based purchases, “the human is no longer present in the purchasing process,” which without intervention “introduces a lot of risk to merchants.” Worldpay is working with Trulioo to authenticate AI agents, verify card credentials, and confirm intent.
- Scale of adoption. Consulting firm Edgar Dunn & Co projects AI agents will handle $136 billion in consumer-to-business transactions this year, climbing to $1.7 trillion by 2030.
Fraud Experts Flag Risks
At the same time, the AI innovators acknowledged the potential fraud risks.
- Breaking assumptions. Google itself acknowledged: “The rise of autonomous agents and their ability to initiate a payment breaks this fundamental assumption and raises critical questions” about security and accountability.
- Good bots vs. bad bots. Turner warned that distinguishing legitimate agentic activity from fraud is essential: “Without significant intervention by the various payments ecosystem players … that introduces a lot of risk to merchants.”
- Authentication gaps. While AP2’s intent mandates and signatures represent progress, Turner said, “there’s a lot of innovation” still required to manage fraud and chargebacks in agentic commerce.
My View: Stop Fraud at Time of Purchase
In my judgment, fraud prevention in the AI era must happen in real time. Waiting until after authorization to detect and remediate fraud is no longer enough.
That means:
- Real-time identity and intent validation. Every transaction must confirm who is authorizing it and under what conditions – a “person present” transaction.
- Agent verification. AI shoppers must be tied to verifiable credentials and resistant to spoofing.
- Ecosystem collaboration. Issuers, merchants, and networks must adopt standards like AP2 broadly and consistently.
- Low-risk pilots first. As Worldpay suggests, testing with groceries, staples, and inventory replenishment makes sense before scaling to high-value categories.
AI commerce can transform global payments. But if fraud prevention lags, criminals will exploit AI or agentic payments at machine speed.
One last thought. Imagine trying to deal with a catastrophe of chargebacks resulting from agentic shopping, false declines and friendly fraud?
To protect consumers, merchants, and issuers, we must embed fraud prevention into the transaction itself. At EQUE, we believe this truth is non-negotiable: stop fraud at the time of purchase, or you risk accelerating it.
Recent PaymentsNEXT news:
Attacking ERP Systems: The Next Frontier in B2B Payment Fraud