Stripe Deploys New AI Tools for Fraud Prevention
Stripe has upgraded its fraud prevention platform with new AI-powered tools designed to provide more accurate, real-time threat detection for its payment customers. The update is intended to improve frictionless onboarding, dynamic risk scoring, and adaptive security workflows. A related guide for engineers details the practical complexities of integrating Stripe into a two-sided marketplace.
- The global cost of online payment fraud for businesses is estimated to be over $20 billion annually, a figure that has risen with the growth of e-commerce. - Stripe's AI-based fraud prevention system, called Radar, is trained on transaction data from the hundreds of billions of dollars in payments processed across the Stripe network each year. - The company recently upgraded its AI architecture from an ensemble model composed of XGBoost and a deep neural network (DNN) to a pure DNN-only model, which has significantly improved performance with their large dataset. - This new "Payments Foundation Model" has improved the detection of sophisticated card testing attacks by 64% for large users in a very short amount of time. - To further enhance its capabilities, Stripe acquired Bouncer, a startup specializing in card scanning and authentication technology to verify the authenticity of a physical card in under a second. - Stripe has also expanded Radar's AI fraud detection to cover non-card payment methods like ACH and SEPA, addressing a 40% increase in volume for these payment types. - The broader market for AI in fraud detection is projected to grow significantly, with one report estimating it will reach nearly $120 billion by 2034, up from $15.6 billion in 2025. - The new AI tools can now combine machine learning risk scores with real-time responses from card issuers, allowing for more dynamic decision-making that can reduce the number of legitimate transactions that are incorrectly blocked.
