Fraudsters have research and development budgets now.
We recently published The Fraud AI Stack Is Scaling Faster Than You Are, which walked through what our data is showing on AI subscription fraud. We found nearly 34,000 disputes tied to major AI platforms since January 2025, and paired with other AI tools on the same claim, it looked a lot like an organized operation. I want to pick up where that piece left off, because the “what” is only half the story. An interesting question is how we saw it at all, and why almost no other data set in this industry could have surfaced it the way ours did.
The compounding effect
The thing about shared data is that it compounds. The more contributors you have feeding a data set, the more accurate your models become and the more accurate your predictions get as a result. That’s a network effect, and it builds on itself over time. The more institutions on board, the more effective the network is for everybody on it.
That’s a familiar idea in identity verification. Consortium data is common practice for account opening, new-customer screening, phone ownership checks — the front door of the relationship. It’s far less common on the post-intake side of things, in the actual dispute resolution process. That gap is what uniquely positions Quavo in this space. What the identity world has been doing at onboarding for years, we’re doing for fraud and disputes.
What that view catches
The most interesting thing our cross-network visibility gives us is the ability to see abuse across entities that no single institution could ever see on its own. Is a customer abusing multiple institutions at once? Is there a fraud farm operating out of one address, filing across issuers who have no idea they’re looking at the same bad actor? That’s the kind of intelligence that lets an institution actually hone in on the people running the dispute process for profit, instead of chasing shadows in their own book.
The AI subscription pattern is a good recent example. It wasn’t just one AI tool showing up on claims. We found a whole stack of AI subscriptions disputed alongside crypto wallets and other AI identity tools like voice or video editing services. Fraudsters need compute capacity the same way any legitimate business does now, and buying that compute with stolen card numbers is just the current version of an old playbook. What made it visible was that the pattern showed up across a limited number of issuers first, and because we could see across the network, we could tell the rest of the network what was coming before it hit them.
It doesn’t stop at a warning
This isn’t just a heads-up to go and adjust your team. Insights are built into the intelligence itself. We’ve started building merchant classifier models that bucket merchants: Is this a scam merchant? Is this one of those social-media-trendy operations that never actually delivers a product? Or is this a legitimate merchant that got hit by fraudsters targeting them?
We’re doing the same thing with high-risk industries like crypto, betting, fantasy leagues, and similar categories that are more attractive to fraudsters by nature. Bucketing those merchants and feeding that into our models creates a feedback loop: the models get smarter about which merchants and which patterns deserve more scrutiny, and that shows up as more predictive scores downstream.
To the risk officer who’s still reluctant
Data sharing sounds like it should make a risk officer nervous, and in a lot of contexts, it should. But fraud is one of the unique areas in financial services where sharing is genuinely the right call. Step back far enough and look at what we’re fighting. Some of this ends up funding human trafficking, terroristic financing, the worst things you can imagine from a humanitarian standpoint. Regulators have recognized that, which is exactly why they’ve given vendors and issuers the room to work together on it. We’re given a real allowance because the stakes are that serious.
And honestly, we’ve barely scratched the surface of what financial institutions and organizations like Quavo can do together. We’re at the tip of the iceberg.
The real competition
There’s a framing I’ve heard floated internally that I think gets it exactly right: this was never supposed to be institutions competing against each other. It’s financial institutions competing against fraudsters. Those are the real villains. Every dollar of intelligence we don’t share is a dollar of advantage we’re handing to the other side.
