How Do You Detect AI Agents on Your Website? Arkose Labs CEO Kevin Gosschalk at RSAC 2026

It's been a long time. But I know the company has been growing. Yeah. Quite a bit.

We're very fortunate to work with really incredible companies and, I think more importantly, we get to solve really fun problems. So we're always in the way of bad guys trying to get through us to attack our customers. That really puts us in the front line of figuring out what do we need to build to disrupt that, prevent that, and all that kind of stuff.

I think RSA reinvents itself every other year with like, what's the big trend? What's the big threat vector we're worried about? Like it certainly changes and ebbs and flows. There is the bingo card that you have. Yeah, yeah. It's interesting now that I've come here so many times since I started the company, but.

This year, I feel like it's like reinventing the category we've been in for so long. So for me it's like really quite a lot of fun. Everyone, once again, is now talking about bots. Bots are gonna be buying things, bots are gonna be going and signing up for stuff. Bots are gonna be writing code or whatever it may be.

But it's agents at the end of the day, autonomous systems doing things for users, which that's really the core of what we've been doing for so long now.

It's also incredibly important from an anti-adversarial mindset because all the bad actors are absolutely looking at this technology and thinking, how can I jerry-rig generating synthetic identity and going, enrolling for a bunch of bank account credit cards and reaping promo fraud bonuses and all that stuff, like that's already absolutely being discussed.

Mm-hmm. And the way these tools interact with the internet is the same way bots have been interacting with the internet for many years. So it's really the same kind of stuff, but with a little bit more intelligence piloting it as opposed to a human having to do that now. Right.

But the concept of a bot going to your login screen or your account application creation screen, or using your app or buying something. But that's always been like a no-no. Like you don't want bots doing that.

Like if you're gonna use those things, you have to do it through an app. You gotta go to the web browser.

But now that's different. So we're gonna keep seeing the bad stuff, trying to do those things, but there's absolutely also gonna be these good agents, agentic commerce, where you want to enable that.

People are gonna be like, to the app, hey, I want to buy a pair of shoes. Go find me the right price. This is my price. The right size. This is the size, price, color, this is where I live. I'm only happy with X shipping time frame. Go find it and it will do that, right? That's a sale you'd miss out on if you block bots.

So I think it's becoming really interesting 'cause now, for us, our customers want the ability to enforce the policies that they want as opposed to just a blanket like, hey, block all the bots. Right. So let's talk about this. Let's name, again, the elephant in the room. Agentic AI.

Yeah. So how do you identify an agent? How do you give them an ID? And then decide, you're good, or nope, you're not. It's complicated. No surprise, as you'd expect. So there are agents who are self-identifying their agents.

So when they connect to your website, in their URL or in their user agent equivalent, they'll add some information in there and you can look that up in a database and say, okay, this is owned by OpenAI, or this is owned by Anthropic. So that's like an easy part of the problem. In certain scenarios they just self-identify.

That is almost 0% of what's actually occurring right now, though the vast majority of agentic traffic is turning up, pretending to be a normal Chrome browser.

One is the behavior of how it's controlling your Chrome and what it's doing. There are signals absolutely today, there's a lot of signals you can collect on, like behavior biometrics, to see the answer to that.

The other thing is, what's it doing? What's the intent? If it's going through and doing stuff in a normal user way, you probably categorize that as a good agent. If it's going in and signing up for an account and then bailing once it gets a promo credit, you probably put that more in the, okay, this is a behavior we don't want.

So this concept of intent is also important. We, for instance, with the Arkose product, have intents around payment fraud intent, fake account creation intent, account compromise intent. And there's a number of signals that we look at to determine someone's trying to compromise an account, someone's trying to phish this user, this is a fake account that's signing up. Those kind of things.

So the biggest irony, I think everyone's like, the concept of a CAPTCHA is meant to be something machines can't solve, which I think that concept is not achievable. Like anything a human can do, you can train the machine to do.

What we focus on is economic disruption. So if we can make the effort and cost of what they're trying to achieve higher than their profit reward, that is the entire strategy we take. So we have technology because they'll go with the lower-hanging fruit.

We have challenge mechanisms which are designed to be expensive to spend time on, human labor attacks or machine vision attacks.

There are no agentic AI systems that can get through everything that they've never seen before.

So the biggest solution, if you want to defeat a large multimodal model, is you show it something novel that's not been trained on, and it doesn't know what to do. 'Cause at the end of the day, yes, we are highly emphasizing it's AI, but really it's predictive text modeling. It's not intelligent.

It's always going to be an arms race. There is no silver bullet in security. As long as there is an incentive to do something, people will be trying to achieve that outcome.

The economics of fraud are really good, quite heavily in their favor. There's a lot of things to attack. They get the advantage of people not knowing when they're going to attack or what they're going after. There's an incredible amount of ways of making money by doing bad things on the internet, unfortunately.

There's one attack that costs companies we work with tens of millions of dollars, to the point where they're not profitable in certain countries they operate in because of this attack vector. You'd be like, that's surprising, I've never heard of that, but I can see why that makes sense.

So when you sign up to a website, sometimes they ask you to put a phone number in, to text you a verification code, as part of the signup flow, or when you're logging in they might text you, like 2FA. It turns out in a lot of countries in the world, like Egypt, Ukraine, and many places in Europe, the cost to send that text message is pretty expensive. It could cost up to 30 cents to text someone in Egypt or Vietnam, like 29 cents to send that OTP code to them, 'cause you've gotta use the local telecom infrastructure.

The other thing with this attack is the telecom providers have premium phone numbers where the person getting the text message receives a small percentage of the cost the sender pays. They use this for game shows, like on TV, text this number, it's gonna cost you a dollar, the game show keeps 80 cents. So what bad actors do is basically a referral fee, an affiliate fee.

So bad actors get access to millions of premium phone numbers, then they use a bot to sign up for millions of accounts and make 3 to 6 cents every time they send an OTP message. Companies are losing tens of millions of dollars to this.

You wouldn't even think of that. And it's not a type of fraud that hits consumers, 'cause it's attacking the platform itself, so consumers aren't even aware it's probably happening. But they're very creative in finding ways to make money, anything they can possibly make money on.

I think agentic technology will resurface things that aren't profitable fraud schemes right now, 'cause you have to have a human in the loop to do it, or something's expensive about it. Now it can be done at scale and the cost has just plummeted. So we're gonna see a reinvention of what was old fraud. It's most likely gonna come back with an agentic flavor on it at much higher scale.