Sardine Launches Dedicated AI Research Lab and $375,000 in Research Fellowships to Advance Fraud and Financial Crime Prevention

• Sardine AI Labs will focus on production-grade fraud and financial crime AI models that meet real world latency, governance, and explainability requirements
• Early results from its foundational AI model for card transaction fraud showed a 68% improvement in fraud detection accuracy for a consumer card issuer, and a 41% improvement for a business card issuer

SAN FRANCISCO--(BUSINESS WIRE)--Sardine, the leading agentic risk platform for fighting fraud and financial crime, today launched Sardine AI Labs, an applied research group advancing frontier intelligence to fight financial crime. The team will study how AI can learn the language of financial behavior, anticipate new attacks and help risk teams make better decisions. Alongside the launch, Sardine is opening applications for up to 5 research fellowships, awarded to independent researchers working to advance the lab's early results.

The next wave of foundation-model innovation is expected to come from domain-specific AI neolabs with access to proprietary, real-world data. Fraud and financial crime prevention is a prime example: even leading academic researchers often lack the large-scale datasets needed to advance the field.

Sardine AI Labs is addressing this gap by tackling some of the most complex challenges in fraud and financial crime prevention. Much like any language, payment activity and financial access behavior contain distinct patterns. Transaction histories, devices, and IP addresses form behavioral profiles that foundation models can analyze to identify anomalies and detect potential fraud.

“We are uniquely positioned to train a model purpose-built for risk because we sit on the industry’s fastest growing fraud and fincrime dataset, which spans more than 6.5B devices, 441M consumers, 3.4M businesses, 6.6 billion transactions and $1.8 trillion in payments," said Soups Ranjan, CEO and Co-founder of Sardine. “Sardine AI Labs will focus on production-grade models that meet real world latency, governance, and explainability requirements.”

Solving the Hardest Challenges in Financial Crime

Financial crime prevention remains costly for financial institutions, contributing to higher payment-processing expenses across card, ACH, and wire transactions. Foundation models have the potential to address persistent challenges that drive these costs.

One example is sanctions screening. Consumers with common names or identities similar to those on sanctions lists can be incorrectly flagged, delaying onboarding for weeks while financial institutions conduct manual reviews. Accurately determining whether two records refer to the same individual remains a complex challenge.

Another is detecting money laundering, terrorist financing, and transactions linked to drug trafficking among the billions of payments processed by banks. Existing systems can generate false positives that delay legitimate transactions while still failing to identify illicit activity, exposing financial institutions to significant regulatory and financial consequences. Foundation models could improve the accuracy and efficiency of both processes.

Sardine AI Labs will explore AI research problems around modeling very large sequences of transaction and user behavior events, transfer learning, adversarial robustness, and explainability. The goal is to turn advances in AI into measurable protection against financial crime. Grounded in Sardine’s network of device, identity, behavioral, and transaction data, the lab is developing models that can understand complex financial behavior and detect attacks they were never explicitly trained to find.

Early Findings Show Stronger Fraud Detection Across Issuers

Sardine AI Labs also published early results from a foundation transformer-based model trained to learn patterns across complete cardholder transaction histories. The card model was trained on about a billion transactions over the last 2 years from over a dozen card issuers.

One of the hardest fraud prevention challenges for new card issuers is the cold-start problem. At launch, these issuers lack the training data needed to detect fraud reliably, yet they are often among the first targets of fraud rings seeking to exploit newly released financial products.

Sardine AI Labs’ new card foundation model is designed to address this challenge. In testing with issuers that were entirely excluded from the model’s training data, the model improved fraud detection accuracy by 68% for a consumer card issuer and 41% for a business card issuer compared with standard machine-learning approaches.

The accuracy gains were consistent across consumer, business, and global cross-border card issuers, demonstrating the model’s ability to generalize across different types of card programs.

“The most important finding is that foundation models can learn directly from a user’s transaction behavior, allowing us to identify fraud more accurately than approaches that reduce that behavior to a set of tabular features,” said Niranjan Shetty, Head of Data Science at Sardine. “Because these patterns transfer across financial institutions, the model is not limited to a single card program. The next step is to make this intelligence fast, explainable, and reliable enough to support real-world risk decisions.”

“The frontier AI labs have shown the immense benefits of having multi-modal AI models which allow you to extract intelligence across different modalities of audio, video and text. We think that the next unlock in fraud and fincrime prevention would occur from creating a multi-modal AI that goes across device, identity, behavioral, and transaction data. This is what excites me the most about the potential ahead with our AI Lab and I can’t wait to see what our research fellows build with us,” said Soups Ranjan, CEO and Co-founder of Sardine.

The model architecture, training methodology, and full results are available in a new technical whitepaper.

Introducing the Sardine AI Fellowship Program

Sardine invites independent researchers currently enrolled in universities in the US or Canada to build on these findings through its new AI Fellowship Program. The program will select up to five fellows to pursue research into how models can:

• Generalize across card issuers
• Learn from multiple streams of financial activity
• Improve as data and model size scale
• Perform with limited customer history
• Improve identity matching
• Detect money-laundering patterns
• Support real-time inference in hundreds of milliseconds
• Extend to areas beyond fraud

Applications are open through December 13, 2026. Shortlisted applicants will be notified in December, and the inaugural cohort of fellows will be announced in mid-January 2027. To learn more and apply, visit sardine.ai/ai-labs.

About Sardine

Sardine is the leading agentic risk platform for fighting financial crime. Our integrated fraud and financial crime solutions unify data across risk teams, enabling real-time fraud detection and automated compliance operations. More than 500 global enterprises rely on Sardine to secure and grow their products, including category leaders across banking, fintech, wealth and retirement, HR and payroll, payments, and software. Customers include FIS, Experian, National Bank of Canada, Nubank, GoDaddy, Deel, Gusto, Paylocity, Xero, and ZoomInfo. Learn more at sardine.ai.