00:00This vulnerability seems to be as a result of user error, not necessarily an attack.
00:05How do you look at that in the context of overall vulnerabilities when it comes to the threat landscape today?
00:11Well, no question. Whenever you lose source code, there's a reputational hit.
00:15But that's, you know, when I looked at this case, they didn't lose customer data.
00:19They didn't lose customer credentials. They didn't lose customer model weights.
00:24And that's where the magic happens when you use agents, because agents have total recall on your data.
00:30And how you purpose those agents. So this is one of those things that happens to companies.
00:35Whenever you have a human in the loop, you have a process that's potentially frail or degrades over time.
00:41So they're going to have to manage through the reputational hit on this.
00:45But at the same time frame, I'm a cybersecurity expert, and I look at this.
00:48Whenever you lose source code, there will be folks that scour that looking at ways to how does this software
00:55work?
00:55And how do we take advantage of it? Maybe in this case with prompt injection.
00:59So all you're really going to get out of this, Tim, is probably a faster release cycle.
01:04Because now you have millions of people looking at your code that you never intended to have them look at
01:08it.
01:08So you're going to have a fast release cycle should there be any identified vulnerabilities from the extra million sets
01:15of eyes that are looking at your source code.
01:17You were the founder and CEO of Mandiant before it was sold to Google Cloud.
01:21You're former CEO of FireEye. You have been doing this for decades.
01:25This era that we live in right now, with soon to be everybody having an AI agent that can act
01:31on behalf of them,
01:32and also seemingly perhaps bad actors having AI agents that can act on them.
01:39How does that look out there today compared with other times in this industry?
01:43Are we more vulnerable now than ever before?
01:46Yeah, that's a great question, Tim.
01:48I think we're going through an inflection point.
01:50And I hate that term, but it's just a dead reality.
01:52Whenever you see a shift change in technology, the unfortunate reality is that usually malintent can be implemented faster than
02:00positive intent.
02:01That just is what happens.
02:02So you look at the Internet, it allows people to commit crimes or try to make money from 9,000
02:08miles away from where they perpetrate the crimes.
02:10And wherever money goes, crime follows.
02:12Wherever information goes, espionage follows.
02:15That's our systems.
02:16So you're going to see in under two years, AI agents, because they can think and learn and have total
02:22recall,
02:23they will become the offense in the cyber domain.
02:26And the cyber domain's always been contested, Tim.
02:29There's always been bad guys on the Internet.
02:32But now these bad guys or the espionage actors are going to be able to use agents to work at
02:38compute speed.
02:39So the change we're going to live in the cyber domain is that you're going to see the speed of
02:44attack go way up.
02:46So vulnerability discovery will be condensed down to seconds rather than days or minutes.
02:51And you're going to see over time defense have to be autonomous.
02:54It can't have humans in the loop.
02:56So we're going to go through that inflection point over the next two years.
03:00But agentic AI is absolutely going to speed up the risk that everybody has in the cyber domain.
03:06Kevin, forgive me.
03:07We only have 20 seconds left.
03:08But how are you now doing that and addressing that at Armadan?
03:11Well, we are addressing it.
03:13What you're going to see is Armadan is going to be all offense all the time.
03:17We're going to be the agents that train your defensive agents so you can respond at the speed of compute
03:22and respond autonomously to the threats that emerge.
03:25New and novel threats will be found by the good guys before the bad guys.
Comments