0:00
AI now writes attacks that get past filters about half the time and get clicked about 60% of the time, because they are so convincing.
Email has been plagued by fraudsters since the very first message was sent, and it's only getting worse.
Cy Khormaee, 0:00
0:55
Instead of people writing rules, task-specific models trained by AegisAI work as a team, reasoning and calling one another to investigate every message.
And these models work as a team. They reason and call one another
Cy Khormaee, 0:55
1:24
Yes. In the Robinhood case, agents spotted subtle template changes, then sent an LLM-powered browser to open the links first and find the credential theft behind them.
One of the hardest things that we can detect is actually attacks coming from trusted vendors.
Cy Khormaee, 1:24
2:55
More email means more varied data, which means better detection, which earns the right to see more email. The same agents and data then extend to insider threats and data leaks.
This simple data flywheel enables us to compound our advantage
Cy Khormaee, 2:55
3:22
Attackers watch for funding rounds, new hires and big deals. Most teams invest in security only after an attack.
Teams invest in security too late after the attack.
Cy Khormaee, 3:22
Cy's four-minute talk: why AI-written phishing now gets past both filters and people, how AegisAI's team of agents investigates every email, and why security has to start before the next raise, hire or deal.
Cy Khormaee, CEO and co-founder, AegisAI
Machine transcript from Riverside, lightly corrected.
0:00 Cy Khormaee: Hey everyone, wow, you're gonna get the full cybersecurity hour here, so there's no no stopping it. So just to introduce myself, I'm Cy, I'm the CEO and founder of AegisAI, and I'm here to save your inbox. So before I get started, let me ask: who here has seen scams or spam in their inbox before? Everybody? See all the hands going up? Excellent, you get the problem. Email has been plagued by fraudsters since the very first message was sent, and it's only getting worse.
0:26 Cy Khormaee: AI is creating a new wave of sophisticated attacks that bypass all existing controls, including you, the user. We're seeing about 50% of these emails bypass your email filters, and then about 60% of you all clicking on them just because they're so compelling and so sophisticated. That's why we built AegisAI. And you're gonna hear this term a lot: fight AI with AI. You just heard it in the last talk. So we were the team at Google that secured Gmail, Chrome, and ads for the last decade. And we learned that the old way, the way of using teams of humans,
0:55 Cy Khormaee: seeing attacks and writing rules just was not going to scale in this new era of AI. So we built a new team, a team of agents, and we've custom trained our own task-specific models that are able to detect threats that we see in email, like fake documents, credentials thefts, invoice scams. And it was only through training our own models that we got the speed, precision, reliability to really scale this to be effective at the enterprise level. And these models work as a team. They reason and call one another
1:24 Cy Khormaee: to investigate every single attack. And let me show you how. Cue video. There we go. One of the hardest things that we can detect is actually attacks coming from trusted vendors. In this case, Robinhood was compromised and sending malicious emails to its customers. The attacks coming from inside the house. So what Aegis is able to do is able to read the email, look for very subtle differences. For example, even just subtle changes in HTML templates that look suspicious and use that to trigger deeper dives. For example, agentic sandboxes. We have an LLM powered browser that can go out there and
2:00 Cy Khormaee: step on landmines before the user does to spot things like credential theft and phishing or these fake CAPTCHAs. And those are really the smoking guns that we look for, and that's the power of Aegis. We're able to use these agents to recursively investigate every single threat, every single message in your inbox to find that smoking gun. And we go back to the slides.
2:22 Cy Khormaee: The other way? There we go. The efficacy of the solutions, we're able to find about 22% more attacks than other vendors at a 90% higher accuracy. We're able to use our agents to cover the full breadth of the email solution from phishing to spam to DLP to training. This has won us the trust of some of the top AI innovators in the space, like [unclear], Notion, LangChain, and Reflection AI. So now that we've built this team of agents, really kind of the strategy is just to keep on growing it.
2:55 Cy Khormaee: As we see more email, we get a better diversity of data, which enables us to have better detection, and that earns us the right to look at more email. This simple data flywheel enables us to compound our advantage to become one of the enduring leaders in the space. And this gets us just one step closer to doing what really we want to do in our mission in life, which is to protect you, all, the users. We're rapidly expanding our team of agents to cover new areas, like looking at insider threats or data leaks.
3:22 Cy Khormaee: And we're doing it all with the same harnesses, the same data, the same flywheel we've used for the core email phishing product. And so for this room, since I know you're probably not all security experts, although you will be at the end of this hour, here's why you should care. When you raise your next round, you hire your next dozen employees, or you close that next big deal, the bad guys will be watching and they will come for you. We've seen it a thousand times. Teams invest in security too late after the attack. So just remember 90% of these attacks start with email.
3:52 Cy Khormaee: So do yourselves, do us a favor, and think about security before you get hacked so you can keep on building. And so my one call to action is come check us out at aegisai.com and I'll show you how we can secure your inbox. Thank you.