Feedzai has unveiled RiskFM (Risk Foundation Model), which it calls the industry’s first tabular foundation model built for financial data and risk decisioning, the company said on 24 March 2026.
RiskFM spans fraud detection, anti-money laundering and broader risk decisions across the financial crime lifecycle, replacing rules and machine learning models built one customer at a time.
Feedzai says the model is trained on a broad, global dataset covering onboarding, digital activity, payments, transfers and AML workflows, unlike approaches limited to card network data.
The company argues that transactional data is harder to model than language, which is constrained by grammar, so a model purpose-built for tabular data is needed.
For banks, a pre-trained risk model could shorten the time to deploy fraud and AML detection for new products and markets.