Plaid launches 'Protect' anti-fraud product as financial fraud rises 25% a year

happy place. Book a wander with inspiring views, hotel grade aes, dreamy beds, top tier cleaning, 24/7 concier service. It's wander allows you to go golden retriever mode because you just don't have to think about anything. You just be chasing the ball, chasing the beach. Uh, well, we have Zach again from Plaid.

uh third time on the show. Welcome to the stream. Great to see you. It's always a pleasure. I'm I'm always blown away when you join because uh it's always so much fun talking to you. So, thanks so much for joining. Uh great to have you. What's What's the latest in your world? Uh well, thank you so much for having me.

I'm I'm I'm honored to be here and uh it's always fun. So much fun. Uh latest in our world is uh we did Plat Effects today. This is our our annual customer conference where we announced a bunch of the stuff that we've been working on for the past year. Um it's been uh super fun.

And the biggest announcement was a product called Protect, which is this uh amazing cutting edge anti-fraud product. Looks at all the data across all the all all of the usage and the data flowing through the cloud network and uh builds a really great anti- fraud tool. Yeah, that's that's the headline of the day.

H how how oneshottable is fraud these days? Can you just pipe a ton of customer data into like a single context window and just ask and can the prompt just be like is this fraud now or or is it is it still like leveraging all the traditional fraud tools? Yeah, I mean the the the hard part is twofold.

Okay, first is you have to put all of your historical data in it to say, hey, is this pattern anomalous? Is something weird going on here? So the question is like can you can you get all of your data into that?

And in plaid's case, can you take all the data that we see across all of the fintic apps and all the users and say, hey, what are the patterns that are occurring consistently? So for example, um you know, if you sign up uh for for Coinbase today, that might be a totally normal signup.

you signed up for Lending Club today and you just signed up for six other lending products today. Then that's obviously a potentially fraudulent loan that you're trying to take out. And so like can you can can you get all this data all in one place? That's that's problem number one.

And problem number two is can you analyze it and change the analysis that you're doing at the speed that you need to change it because fraud changes every day. Um some of that is algorithmic. So are the algorithms like chasing the fraudsters in the way that they need to? Some of that is manual.

So we built a tool that actually allows humans to go in and say, "Hey, I heard from, you know, I've seen this thing going on or I heard from some other some other fintech company that this fraud ring is occurring.

Can you search for it and tell me is is that actually occurring to me and then can I build some rules to to protect against it? " So it is a constantly evolving game. Yeah. How how important is this problem?

I feel like every American should just default assume that their information is like floating around the internet available to the highest bidder. So like in in some ways like the only solution is just better tools to prevent this type of activity. Is that is that reality? Mhm. That is the reality.

Um financial fraud is going up around 25% a year right now. Um some of that is or actually a lot of that is AIdriven. So you know people using AI tools to send text messages to convince people to do certain things.

Um, I think, yeah, one answer is absolutely tools and we're trying to build the best-in-class tools and give it to all the fintech companies and frankly then we'll go well beyond fintech with these products. Um, the other is just education like explaining to consumers how it works.

So, you know, my my parents and I we have a a word that we say to make sure that we know that it is the other person on the other other end of the line if we're sending money or frankly doing anything that requires authentication.

Those kinds of little tricks that you can build into your day-to-day processes are always helpful. But it's a combination of tools and processes and and so forth. Yeah, makes sense. Uh can you talk to us about impossible travel problems? I found out about this yesterday.

Uh where where someone's logging in from Singapore 3 hours after they were in San Francisco and they couldn't possibly be there. Is that something that uh that that you're already built a tool around? Is that is that kind of uh like table stakes or or or are there are there like evolutions there? Yeah.

So, I mean, a lot of the the way that the the anti fraud product works, it's called Protect. It's got a bunch of these different analysis analyses that we do. So, there's there's I think just under 20 different analyses that are running now and we're adding more every day. Um, one of them is location tracking.

So, we look at where was the last place that you logged into a fintech app and then if you're trying to do a new thing with a different fintech app on the same user's account in a different place. All right, great. Well, that's that's actually, you know, it's an impossible travel question.

Um, and so that will immediately generate a flag. Um there's a bunch of these other things like we look at stuff like how long has this bank account existed? Turns out a brand new bank account is very likely to commit fraud and an old bank account is less likely to commit fraud.

Um we do stuff same thing for the phone number. How long has this phone number existed? And there's there's a zillion of these analyses that then run into a bunch of statistical work that we do on top of it.

Um but you know every time you look at a fraudster or or an instance of fraud, you can always explain it in retrospect. So you can look back and say, "Oh yeah, obviously, you know, that person was in Singapore and then they were in California 2 hours later. That's impossible.

" But when you're proactively trying to prevent it, like can you actually think through all the possible permutations of weird patterns that might exist? And that's that's that's the complexity that we focus on.

So So uh I imagine at one point this was like very heruristic driven, very uh like a multivariable regression almost just saying like here are the different fraud factors. Let's add up like a risk score. uh have these products moved to a transformer architecture? Are we using LLMs in them?

Are we using modern AI tools or or or are we just kind of continuing to scale up the the the existing uh uh the existing algorithms that have been working for the past few years? Like I don't know how relevant the new models are.

People are kind of like throwing transformers and diffusion and everything right now and I don't know if that's appropriate here. Yeah, I mean the answer is yes to all the above. Um it is not heavily transformerdriven yet in terms of fighting fraud though increasingly um it is becoming that way. Sure.

We definitely use LLM for a bunch of things like autob building tools like doing some of the data cleansing so and so forth on the back. Um but you know fraud is a fascinating industry because um if you talk to any company their goal is to make the fraud not happen on their platform, right?

They want the fraud to be more expensive for a fraudster to do on their platform than it is anywhere else. Sure. Um and so everyone wants to be ahead. They want the new thing. They want like the the new type of fraud solution.

Um but then eventually everything that's new becomes table stakes for everyone because uh you know you kind of have to have it. And so you know we are iterating our way towards more and more complex architecture but uh it's still early days to actually solve.

To me this product feels like something that's just possible with like massive scale. like you probably wanted to build something like this like you know five six seven years ago but but couldn't just because you didn't have like the the sort of broad exposure to all these different touch points. Is that right?

I mean this was the pitch that I was giving for plaid in like 2014. Basically my pitch was when we get big enough we're going to have a lot of data and we're going to build these analytics that financial services doesn't have because uh you know no one can build them and no one has the tech sophistication to do it.

Um, it took us a decade to get to the point where we had enough data to actually build this stuff. And then we acquired Cognto, an ID verification company in 2022, which is a big catalyst.

I mean, in part of the pitch to that team, I was saying, "Look, come use our data, put it together with your data, and then think of all the amazing anti-fraud tools you could build. " So, I mean, this stuff always happens slower than you wanted to, but we couldn't be more excited to get it out there.

Uh h how much of this like like fraud right now, financial fraud, is kind of acting as like a tax on fintech companies and should we be thinking about shifting that tax to the government by giving the government more tools to go after the fraudsters like at the source and just arrest the people that are even trying to do the fraud?

Well, I think one of the big the big challenges is in in crypto a lot of fraudulent activity is happening and and law enforcement is able to trace the activity to somebody in China.

And if you just go to if you go to some if you go to the Chinese government and you say like we know exactly who's doing this, they don't they're not going to act on it. So like it's Yeah. I mean the answer is yes. Yes to all the above. Like the government should be more active in this.

Um, again, much of the, as you said, much of the fraud doesn't stem from people that are inside the US. So, it's not the US government doing it. I'm sure you guys have been following or maybe you've read the articles about pig butchering. Horrible name for a type of fraud.

Like they have these these these humans in fraud rings in like Malaysia with people, right? Yeah. Exactly. And like they're they're often government sponsored or at least government like affiliated in some way. Like there's there's knowledge in the government that that that happens.

And so I mean you know in if that is happening uh great yes we could we could give better data to the Malaysian government but also you got to build the tools. Yeah. Yeah. What about um uh I I don't know how how much you can comment on this.

Um but you know there's been a lot of discussion recently and and frustration and at least after the the Coinbase um uh you know user data issues around like KYC's KYC laws and and effectively the government requiring people to retain all this data.

Do you think that there would ever be any movement uh in DC around really updating these laws to protect to help protect user data or is it just one of those things that that we have to to live with? I mean you you would think they should.

Um, I mean, realistically, yeah, that the concept that um, if if if you're a bank and I sign up for your bank, you have to collect all of my most sensitive data and then you have to store it forever in your database.

On a scale of forever in, you know, many many databases, you have to assume that those databases will be hacked. Thus, you have to assume that on a scale of forever, your data is all going to be public. And so, it doesn't make sense that that that you would retain it.

you should in theory collect it, verify it, um like get say a token uh that it exists, like get some assurance that it exists and then delete it. Um it seems like I I would hope that the regulations change. I think it's it's it's you know regulations move as fast as regulations move.

Um and so I'm not sure that it's going to be uh imminent in that sense. But on the flip side, like the industry is solving it. The industry is building tools that are better and better. You know, relying on a static social security number, sure, yeah, we'll do it.

That's fine if if we have to for regulatory reasons, but the reality is if if if social security number entry is your only uh fraud detection tool, your only fraud avoidance tool, like you're you're in a bad place as it is. Yeah, that makes sense. The industry will solve it basically.

Can you talk uh uh this got kind of lost among other uh uh massive news cycles, but there were some updates to DoddFrank 1033 that a few weeks ago. um you guys um talking offline were were kind of concerned about what that could mean just kind of broadly for consumers.

Any any kind of can you give us kind of an overview of what happened? I know I think when we were in DC you were kind of covering a little bit of this stuff on the show as well. Totally. Yeah, I think this is a a big issue that probably doesn't get as much coverage as it should.

So um in the DoddFrank law um the the the wrote a provision that basically says consumers have access to their financial data and they can assign that access to uh to third parties such as Plaid to act on their behalf. Um and uh then there was never any rule written on that.

It was it was regulatory authority that was given to the CFPB. Um no rule written for a long time under Trump first uh first administration. Um Kathy Craninger, the head of the CFPB at the time, uh started writing the rule for this. It then continued into the Rohood Chopra era of CFPB under Biden.

Um the rule that emerged from that was a very far-reaching huge overstep of a rule. Um and in uh the second Trump administration, now there's a push to either pair back or get rid of the rule.

What actually happened is the day that the rule um was published by Ro Chopra um a group of the banks sued you know that same day meaning they had the lawsuit pre-prepared beforehand and just click submit um and so that that kind of is ongoing um the wrinkle that's come up recently is that the currency has basically said um something effective yes we agree with the banks and we should get rid of the rule but also the way that they're structuring the language in that dismissal may foreclose on some of the regulator things that were set up and could actually put at risk um some of the the ways that open banking works in the United States today.

Meaning um it could it could allow certain banks to put much more pressure on the ecosystem. Um they could try to turn off data in certain areas. They should try to limit data in certain areas. They could more more likely um what's going to happen is they're going to try to limit certain industries.

So if you remember um you know a couple years ago uh there was this big uh debanking thing that went on where the banks were trying to push back and turn off crypto. Yeah. This would give them the ability to turn off crypto.

This would give them the ability to turn off you know gaming or things like that or basically any any industry that they might feel is competitive with them. And it wouldn't be all the banks doing it as a group. It would be each individual bank kind of making their own determination of this.

And so we think that you know we don't mind getting rid of the rule. We just don't want to foreclose on the fact that consumers really should have the ability to choose the financial products that they interact with. Yeah, that makes sense. Uh, did you have a reaction to this the circle IPO?

I I I you know, obviously it performed very well, but and and and stable coins have been in the news. Um, but I'm I'm interested to hear kind of like what the next most exciting uh instantiations of that technology might be.

Uh I was talking to a payroll company CEO and uh he was kind of mentioning that uh yes it's lower fees and faster but interestingly stable coins might allow them to kind of recapture float and then earn yield on that in an interesting way that I wasn't expecting.

So like the the the potential economic benefit to some companies might be massive if they're able to kind of uh custody their own treasuries essentially through stable coins. Um uh what have you been seeing in the stable coin world broadly? What are the interesting knock-on effects?

Um how much excitement is there among stable coins from your partners? Yeah, I mean there there there's a ton. Yeah, there's a lot of amazing companies. I think you talked to pretty Have you already talked to them? You're going to talk? Yeah. Yeah, just an hour ago. Just earlier. Great.

Yeah, I mean huge congrats to them. Like amazing to see what they built and very excited for the ecosystem. Um I think you know from our perspective what's happening in stable coins right now is not what will happen.

So what's happening right now a lot of people are using it as a currency hedge a lot of people are using it for crossber um payouts or or money transfers that otherwise would just be frictionful um and then you know a lot of people are using it uh just as a as a form of a s savings account that has higher yield and while those are all big tams it's not that that interesting relative to the rest of what what what could go on in financial services um I think increasingly we're going to see more and more consumers having the ability to move money easily between say a bank account and a crypto wallet uh holding something in USDC or or or another uh another another stable coin.

Um and yeah, you're right. There there are big opportunities for people to hold uh elements of their treasury or elements of their float and earn meaningful interest off of it.

Um increasingly I think that that that float which was historically held on to by the platforms may well get shared more and more um with uh with the underlying companies. So payroll uh companies if the float was all held at the bank now they're actually able to participate in a portion of the float.

So that's a huge new revenue stream for them while interest rates are what they are I will say. So that'll create increasing volatility in their in their revenue streams because if interest rates go down then obviously the float goes down.

Um so I do think that that's that's a a big potential thing that a lot of uh a lot of companies might want to also offering more uh tender types allows a lot of the large platforms to keep dollars on their platform more.

Um, so you see some of the big payments processors, their goal obviously is to have more float on their platform because they get a slice of that float.

Um, offering more tender types, offering it not just to be held in dollars, but you could hold it in in in in crypto, you could hold it in a stable coin that keeps the dollars on their platform even more, which gives them both more leverage and and more open in the long term. Very cool.

What are you guys doing on the MCP side? I I know you you had some announcements recently there. Yeah, so those that was another big uh part of the announcements today. I mean the short answer is like MTPs make it easier for developers to build.

Um I think the concept of uh a rest API was new when we started plaid because everything was on SOAP. Now the concept of uh you know writing to a raw API without using cursor or winds surf or something like that is increasingly becoming uh rare and will become rare over time.

And so really for us it's just staying on the cutting edge building tools that help developers build faster on flat and then allowing them to get better analytics on their usage their uh their interactions with our system so on and so forth. So it's fun. It's really fun. That's awesome.

Do you have any type of broad thesis around how agent you know I think there's like a lot of narrative around okay agents will just default use crypto assets or or you know any anything around that. Anytime there's sort of too it's coming. Yeah. Anytime there's sort of an accepted narrative.

I I just you know want to kind of push deeper. I'm curious if you have strong opinions. I do not think they will default these crypto assets. Um, I think crypto assets might be easier at first, but um, look, the the bank account is already effectively a digital first product or now is a digital first product.

Did not start that way certainly. Um, cards oddly are not yet fully digital first, but are moving that direction. Like still you have to know a 16-digit card number to actually like authenticate a card.

um whereas bank accounts have ooth already set up and then kind of the next phase of that is obviously crypto as crypto starts uh being digital native and so it becomes easier and easier to authenticate into things.

My take is that all three of those and many other tender types will become available to to to these uh to to to LLM uh and to agents but um it's going to take a little while for the industry to evolve but fast forward two three years and it'll be all the same. Very cool. Anything else Jordy? No, this is great.

This is fantastic. Congrats. I'm curious, did did Chesky tell you to to move to this kind of release cycle or or you've been doing it for a while? Is it sort of the Chesky the Chesy play the Chesky method of entrepreneurship?

We've been doing it for a little while, but man, they have developed such an impressive track record of doing it and all the podcasts that he went on to talk about it, I think everybody else in the world decided to copy it.

So, maybe we were inspired by seeing what they did beforehand, but certainly we've learned a lot from from from from from listening to their playbook. Yeah. the the the the jobs approach, you know, walked so the chesky method could run. Method. Exactly. Um, awesome.

Well, I'm excited to read through the the other announcements later and uh come back on anytime. It's great to congratulations. Cheers. Really quickly, let me tell you about Bezel. Go to getbzzle. com. Your bezel concierge is available now to source you any watch on the planet. Seriously, any watch. I love that.

There was one YC yesterday that had the the Daytona and the and the Texas Timex. Yeah, the day date was the day on JLC reverso. Yeah. Well, when we called the founders on it, they were like, I had a lot of embarrassing. It's not super YC coded to have a to have a hitter on the wrist, but nothing throw 100K into a PC.

You know, work 20%. That's the Cluey method. Absolutely. That is the Cle method. No, build a generational company and then uh buy a nice piece on bezel for your for your father. Father's Day is coming up. Head over to bezel, get your dad a watch.

Um, let's check in with our intern, Tyler, who's pivoting from computer science to art history. What do you got for us, Tyler? Oh, he's at the Museum of Modern Art now. That's fantastic. Yeah.

So, um, earlier I I was thinking like I think before I should really start studying, I should kind of just like benchmark my stats. Okay. So, I took two different tests. Uh, the first one was the it was an AP art history practice exam. Okay.

Um, so, you know, it's like for high schoolers, average high schoolers like pretty dumb. Yep. I thought I would do pretty well. I got 12 out of 29. So that's 41%. So failing failing, you know, some some place to grow there.

And then I also took like, you know, maybe um that exam is just like weirdly, you know, it's like very specific knowledge or something. I took another exam. Um this was the Bratannica art history quiz. Um on that one, I got 24 out of uh 60. So that's uh 40%. So uh you know, I have a lot of room to grow.

Well, there there's about an hour left in the show. So, so were you using cursor for art history to to to answer those questions or were you Yeah, I was not. But that was earlier in the show. I have been studying. Um, so I think I'm going to take it, you know, 20 minutes before the show ends. Okay.

Uh, we'll see how I do them. Okay. Um, but I was also kind of just looking um, you know, at what are the kind of careers that art history majors generally go into. Yes. So it like if you kind of want to stay in the art world, there's basically uh you work in a museum or you work in kind of art dealership gallery. Yeah.

So first I looked at kind of be becoming like an a museum curator. Um for that it's you basically need a PhD. Okay. So I don't think I'll be able to get that before the show ends, but I was looking at art dealership. Um and obviously like for that you kind of need some some funding to start with.

So I took uh producer Ben's ramp card. Okay. Okay. Laid down some art pieces. They're going to come in next couple weeks. Great. There you go. So, yeah, I'm already starting this career. I'm feeling pretty good about myself. You flip you flip those pieces, bake it all back, pay off the ramp card, reward business.

Yeah, we will be checking in with his IRRa, seeing if he can beat the average venture fund. Yeah. Yeah. Just beat the market. Yeah, it's that easy. Um, amazing. Great work, Tyler. Yes. Yes. Making lots of lots of promising uh uh promising uh progress. That's very good.

Um, uh, anthropic co-founder Ben Ma Ben Mann says that we'll know AI is transformative when it passes the economic touring test. This is something I was talking about earlier. Give an AI agent a job for a month. Let the hiring manager choose human or machine when