The financial technology sector witnessed a significant technological leap this week as Feedzai, a global leader in AI-native financial crime prevention, announced the launch of RiskFM, a groundbreaking Risk Foundational Model designed to redefine the parameters of fraud detection and anti-money laundering (AML) efforts. By introducing a Tabular Foundation Model specifically engineered for the complexities of financial data, Feedzai aims to move beyond traditional machine learning constraints, offering financial institutions a more agile, accurate, and scalable defense mechanism against increasingly sophisticated criminal networks.
A Paradigm Shift in Financial Artificial Intelligence
The introduction of RiskFM represents a departure from the industry’s recent preoccupation with Large Language Models (LLMs). While LLMs like GPT-4 have demonstrated remarkable capabilities in processing natural language, audio, and video, their application in the high-stakes world of financial transactions has been limited by the inherent nature of the data. Financial transactions do not follow the same linear causality or grammatical structures found in human language. Instead, they are characterized by "tabular" data—structured rows and columns representing timestamps, merchant IDs, transaction amounts, and geographic locations—that require a fundamentally different architectural approach.
Feedzai’s RiskFM is built to address this specific challenge. Trained on a massive, global dataset that spans the entire lifecycle of financial activity—including customer onboarding, digital sessions, payment flows, and cross-border fund transfers—the model provides a holistic view of risk. Unlike traditional models that are often siloed within specific departments, RiskFM serves as a unified intelligence layer capable of making real-time decisions across various fraud modalities and AML workflows.
The Technical Evolution: From Rules to Foundational Models
To understand the significance of RiskFM, one must look at the evolution of financial crime detection. For decades, banks relied on rule-based systems (e.g., "if a transaction exceeds $10,000 and occurs in a foreign country, flag it"). These were easily bypassed by criminals. The subsequent shift to machine learning brought Gradient Boosting and Deep Learning strategies, which improved accuracy but required intensive "feature engineering"—a manual process where data scientists must define which variables are important for the model to analyze.
RiskFM eliminates much of this manual labor. As a foundational model, it possesses an inherent understanding of financial behavior patterns, allowing it to perform with the high-tuned precision of supervised models without the need for constant human intervention in the feature-definition phase. According to internal benchmarks released by Feedzai, RiskFM consistently outperformed traditional Gradient Boosting and standard Deep Learning strategies in detecting complex anomalies such as mule accounts and sophisticated money laundering schemes.
Chief Science Officer Pedro Bizarro highlighted the unique difficulty of this domain, noting that "next transactions are far less predictable than the next word in a sentence." This unpredictability stems from the fact that financial risk is an "adversarial" domain. Unlike a language model that predicts a word based on established linguistic rules, a risk model must predict the actions of human adversaries who are actively trying to hide their tracks and adapt their strategies in real-time.
Chronology of Innovation and Global Scale
The launch of RiskFM is the culmination of nearly two decades of research and development at Feedzai. Founded in 2008 and headquartered in New York, the company has steadily climbed the ranks of the fintech world. The company first gained significant international attention during its debut at FinovateEurope in 2014, where it showcased early iterations of its real-time processing engine.
Since that debut, Feedzai’s growth has been exponential. Today, the company’s AI-native platform serves as the digital backbone for some of the world’s largest banks, payment networks, and merchant acquirers. The scale of the data fueling RiskFM is staggering:
- Consumer Reach: The platform protects more than one billion consumers globally.
- Event Processing: It processes approximately 90 billion events annually.
- Transaction Volume: Feedzai secures an estimated $9 trillion in payment volume every year.
This vast data lake provided the necessary "fuel" to train RiskFM. By analyzing trillions of dollars in historical transactions across different geographies and institution types, the model has developed a "global intuition" for what constitutes legitimate versus fraudulent behavior.
Strategic Impact on Financial Institutions
For Chief Risk Officers (CROs) and Chief Technology Officers (CTOs), the implications of RiskFM are both operational and economic. One of the primary challenges in modern banking is the "cold start" problem—the difficulty of accurately assessing risk for a new product or in a new market where historical data is sparse. Because RiskFM is a pre-trained foundational model, it can be deployed in new environments with minimal local data, providing immediate protection based on its broader global training.
Furthermore, the model’s ability to operate across multiple institutions simultaneously allows for a "herd immunity" effect. When a new fraud tactic is identified at one institution, the foundational intelligence of RiskFM can be updated to protect all other institutions using the model, significantly reducing the window of opportunity for criminals.
Chief Product Officer Pedro Barata emphasized that this launch is the result of a multi-year strategic investment. "RiskFM proves our multi-year investment in foundation models is paying off," Barata stated. "We’re not just part of the conversation; we’re defining how it applies to the complexities of global financial crime prevention."
Addressing the Global Crisis of Financial Crime
The release of RiskFM comes at a critical juncture for the global economy. According to estimates from the United Nations Office on Drugs and Crime (UNODC), between 2% and 5% of global GDP is laundered annually—amounting to as much as $2 trillion. Meanwhile, consumer fraud, fueled by the rise of instant payment systems and social engineering, has reached record highs.
Regulatory bodies such as the Financial Action Task Force (FATF) have increasingly pressured financial institutions to adopt more robust, technology-driven AML and Know Your Customer (KYC) protocols. In this regulatory environment, traditional systems often produce a high volume of "false positives," which frustrate legitimate customers and overwhelm compliance teams. RiskFM’s superior accuracy promises to reduce these false positives, allowing banks to focus their investigative resources on truly high-risk activities.
Recognition and Future Outlook
The industry’s reception of Feedzai’s innovation was solidified shortly after the RiskFM announcement when the company was named to Fast Company’s "World’s Most Innovative Companies 2026" list. This accolade places Feedzai among a select group of enterprises recognized for moving the needle in their respective fields through original research and practical application.
Nuno Sebastiao, Co-Founder and CEO of Feedzai, expressed the company’s broader vision following the recognition: "We at Feedzai are honored by this prestigious recognition of our innovation and research in trusted AI to build a world of safer money." Sebastiao has long advocated for "Trusted AI," a framework that emphasizes transparency, fairness, and accountability in algorithmic decision-making—principles that are integrated into the RiskFM architecture to ensure compliance with emerging AI regulations such as the EU AI Act.
As financial transactions continue to move toward real-time, 24/7 processing, the window for detecting fraud is shrinking from minutes to milliseconds. The launch of RiskFM suggests that the future of financial security lies not in static rules, but in dynamic, foundational intelligence layers that can learn, adapt, and scale at the speed of global commerce. By bridging the gap between advanced AI research and the practical realities of the financial sector, Feedzai has positioned RiskFM as a vital utility for the modern digital economy, potentially setting a new standard for how the world’s money is protected.
