← Back to archive

How Stripe Uses AI for Fraud Detection

Love how Stripe uses AI not just for hype, but to solve real, messy internal problems.

They built a transformer-based foundation model for payments - think of it like an LLM, but instead of language, it's trained on billions of transactions.

A single vector per transaction that captures deep semantic meaning. Like words in a sentence, payments have context, patterns, and relationships.

This boosted Stripe’s detection rate for complex card-testing attacks from 59% to 97% overnight. The same embeddings can now power fraud detection, dispute handling.

How Stripe Uses AI for Fraud Detection

Originally posted on X

PS: I'm pretty active on 𝕏 if you'd like to follow along

Archive

2026

2025

2024

2018