Marqeta Deploys AI to Combat Surging, Sophisticated Payment Fraud

OAKLAND, CA – March 31, 2026 – As financial criminals increasingly weaponize artificial intelligence, card-issuing platform Marqeta is fighting fire with fire. The company today announced a significant enhancement to its fraud prevention toolkit: an AI-powered risk score designed to analyze transaction threats in real time, directly at the point of authorization.

The new capability, integrated into Marqeta’s Real-Time Decisioning (RTD) product, aims to address two of the most persistent challenges in digital payments: the dramatic rise in sophisticated fraud and the costly frustration of false declines, where legitimate transactions are mistakenly blocked. The move signals a critical shift in the industry, moving beyond static rules-based systems toward dynamic, predictive defenses.

A New Front in the AI Arms Race

The urgency for such advanced tools is underscored by stark industry forecasts. Research from Juniper Research projects that global payment fraud will skyrocket by 153% between 2025 and 2030, driven by the increasing sophistication of criminal enterprises. Fraudsters are no longer just lone actors but are often part of organized networks that leverage generative AI to create convincing deepfakes, synthetic identities, and highly personalized phishing scams at an unprecedented scale.

This new reality renders many traditional fraud prevention methods, which rely on predefined rules and known threat patterns, increasingly ineffective. In response, Marqeta’s new system employs machine learning to stay ahead of these evolving threats. The AI-powered risk score analyzes over 300 distinct real-time transaction attributes—from the time of day and purchase amount to device information and geographical location—and compares them against historical behavioral patterns. This entire complex analysis occurs within the milliseconds it takes for a payment authorization decision to be made.

This combination of Marqeta's existing custom authorization rules with the predictive power of machine learning is designed to continuously identify novel fraud patterns as they emerge, effectively engaging in a technological arms race against agile adversaries.

Beyond the Buzzword: The Power of Proprietary Data

In a market saturated with AI claims, the effectiveness of any machine learning model depends entirely on the quality and relevance of the data it is trained on. This is where Marqeta believes it has a distinct advantage. The new risk score is not a generic, off-the-shelf solution; it is trained on the vast and varied streams of proprietary transaction data flowing through its own platform.

Marqeta powers a diverse range of card programs, from on-demand delivery and travel booking platforms to expense management and modern banking applications. The fraud patterns for a food delivery service are vastly different from those for a corporate travel card. By training its models on this specific, use-case-driven data, the platform can build more nuanced and accurate risk profiles. The system is designed to automatically adapt not only to broad market shifts but also to the unique behaviors of individual cardholders within a specific program.

This data-centric approach allows Marqeta’s customers to move beyond one-size-fits-all fraud rules. Instead, they can run risk scenarios and fine-tune controls based on their actual cardholder data, creating a more tailored and effective defense that is unique to their business model and customer base.

Balancing Security with Growth and User Experience

For any business operating a card program, fraud prevention is a delicate balancing act. Overly aggressive security measures can lead to a high rate of false positives—legitimate transactions incorrectly flagged as fraudulent. These false declines are not just a momentary inconvenience; they erode customer trust, lead to abandoned carts, and represent significant lost revenue. Some industry estimates suggest the cost of false declines can even exceed the cost of actual fraud.

Marqeta's new offering is positioned as a tool to recalibrate this balance. By using AI to generate a more precise risk assessment for each transaction, the system aims to reduce the number of legitimate customers who are blocked, thereby improving the overall user experience and enabling higher transaction approval rates. According to the company, this allows clients to expand their card programs with greater confidence.

“By embedding AI-powered controls and advanced machine learning into the authorization process, we enable customers to expand confidently while also strengthening their fraud defense as they scale,” stated Anthony Peculic, Marqeta’s Interim Chief Product Officer, in the announcement. This focus on business enablement highlights a strategic shift from fraud prevention as a simple cost center to a potential driver of growth and customer satisfaction.

Navigating a Crowded and Complex Landscape

Marqeta is entering a dynamic and competitive field. A host of companies, including established players like Featurespace and Feedzai and specialists like SEON and Alloy, are already leveraging AI and machine learning for fraud detection. These firms utilize everything from behavioral analytics to digital identity networks to provide sophisticated risk management for financial institutions.

Marqeta's strategy appears to be its deep integration within the card-issuing ecosystem, leveraging its unique data position to offer a highly specialized solution for its own clients. However, the widespread adoption of AI in finance also brings significant challenges, particularly concerning data privacy and algorithmic bias. AI models are trained on historical data, and if that data contains hidden biases, the model can perpetuate or even amplify them, potentially leading to discriminatory outcomes.

Furthermore, the complexity of these models can lead to a “black box” problem, where it becomes difficult to explain why a specific transaction was declined. This lack of transparency is a major concern for regulators and consumers alike. Addressing these issues requires robust data governance, a commitment to developing explainable AI (XAI), and maintaining meaningful human oversight. As AI becomes more embedded in financial infrastructure, navigating these ethical and regulatory minefields will be just as critical as technological innovation itself.