Card testing fraud is up industry-wide in 2026. On Stripe, it’s down 80%. That gap didn’t happen by accident — and if you’re a fintech startup or small payment platform still relying on basic rule-based fraud filters, this is what you’re missing.
According to Stripe’s blog, the company deployed its Payments Foundation Model — trained on tens of billions of transactions — and detection rates for sophisticated card testing attacks jumped from 59% to 97% overnight. For fintech founders evaluating fraud infrastructure, that’s the number that matters.
What Stripe’s Payments Foundation Model Actually Does
Traditional fraud models are built for specific tasks: block this card pattern, flag this IP range. Stripe’s Foundation Model works differently. It distills each payment into a single embedding that captures hundreds of signals — not just obvious ones like ZIP code or bank, but subtle behavioral patterns no human analyst could track manually.
The practical result: card testing attacks on large Stripe users dropped 80% while industry-wide attacks increased. According to Stripe, $6 billion in legitimate declined transactions were recovered for users in 2024 through improved retry logic alone.
Three other numbers from Stripe’s announcement that fintech operators need to see:
- Businesses using Authorization Boost see a 2.2% average increase in acceptance rates, with some seeing up to 7%
- Radar users saw dispute rates drop 17% last year even as industry-wide ecommerce fraud rose 15%
- Smart Disputes helped Vimeo and Squarespace recover 13% more chargebacks automatically
What This Means for Fintech Startups Not on Stripe
If your platform processes payments through a different PSP, this announcement still affects your strategy. Stripe is now offering Radar risk scores for payments processed by other PSPs — meaning you can plug Stripe’s fraud intelligence into your existing stack without migrating everything.
DoorDash tested this and saw a 10% reduction in chargeback costs by using Radar scores on transactions both on and off Stripe. For a fintech startup processing $5M monthly, that’s a meaningful number.
The other signal worth watching: Stripe’s fraud prevention now covers ACH and SEPA transactions, not just card payments. According to Stripe, users see a 42% reduction in SEPA fraud and 20% reduction in ACH fraud on average. If your platform handles bank transfers, this closes a gap most small fintech teams leave open.
The One Thing to Configure This Week
If you’re already on Stripe, the highest-leverage action right now is enabling Radar for Fraud Teams with intelligent 3DS triggers. Early users have seen over 30% reduction in fraud on eligible transactions. It’s not on by default — you have to turn it on in your Radar settings.
If you’re evaluating fraud vendors, ask any provider two questions: What is your training data size? And how often do you retrain models? Stripe’s advantage comes from processing data across millions of businesses. A vendor trained on a fraction of that transaction volume will miss novel attack patterns.
Key Takeaways
- Stripe’s Payments Foundation Model cut card testing attacks 80% on its network by jumping detection from 59% to 97% — trained on tens of billions of transactions
- Radar fraud scores are now available for non-Stripe transactions, letting fintech platforms add Stripe’s AI without full migration
- ACH and SEPA fraud protection is now live — average 42% SEPA fraud reduction and 20% ACH fraud reduction for Radar users
Is your current fraud stack configured to handle novel card testing patterns — or is it only catching the attacks it’s already seen before?
Source: Stripe Blog