JK

Jason Kikta

Things Jason Says on Podcasts

Where to Find Them

Jason Kikta has been a guest on CyberWire Daily , Irregular Warfare Podcast and Risky Business .

Recently: “Risky Business #860 -- Google: Agentic hacking has “crossed a threshold”” on Risky Business (November 2026); “Not every headhunter is hiring.” on CyberWire Daily (June 2026); “The Digital Bear in Ukraine: Russian Cyber Operations Since 2014” on Irregular Warfare Podcast (April 2023).

What They Said

“Who this really affects is going to be the smaller software vendors and the open source projects who just don't have the resources. Cool, I can get free model access for six months. That's neat for six months, but what do I do after that? And a lot of these AI vulnerabilities, yes, it is technically a vulnerability on paper, but there is no actual path to target, or exploitation is extremely hard, or relies on eight different corner cases lining up. Is that really urgent to patch in 30 days?” — Jason Kikta, Risky Business

Patrick asked whether Anthropic and Google tightening their vulnerability disclosure windows to 30 days will help. Jason says it pressures exhausted maintainers and small vendors, and that many AI-found bugs aren't practically exploitable.

Risky Business · 2026-11-18 Permalink → Listen →
Risky Business Around 36:22 into the episode
James Wilson

Yeah, I too know a lot. What will you pay me? Yeah,

Patrick Gray

…does this, does this help? Does it help? What do you think, Jason? Is this a good motivator to get people to get their ducks in a row and auto-generate patches and spin up a modern program? Or is it just like it just, to me, it just feels like a case of what the Americans say. It is what it is.

Jason Kikta

It feels a little bit like, you know, getting kicked wire down, right? And that's another thing on all these outages we've seen is like everyone's exhausted. Everyone's beyond exhausted. And, you know, having somebody pressure you harder on bugs, you know, and again, we're not talking about the massive multinational companies here that are, you know, in a trusted program. Like who this really affects is going to be the smaller software vendors, the open source projects who just don't have the resources. And like, cool, I can get, you know, free model access for six months. Like that's neat for the six months, but what do I do after that? Right. You know, right now, you know, there needs to be a lot more thought on how, you know, how to help people dig out of this rather than watch my new whiz-bang model go off and discover a bunch of things. And it's also, you know, like it's a really limited use case where this would potentially be a useful response simply because, you know, if you need something where a non-US model, you know, frontier model can find the vulnerability during an embargo period and then use it for quiet exploitation and on and on it goes. And oh, by the way, you know, not for nothing, but a lot of these AI vulnerabilities we've seen, like, yes, it is…

Patrick Gray

Oh, now here's some fun news. Google has mathematically like for has done a has done a formal verification on parts of Signal. Why I find this funny is they have formally verified bits of the protocol that absolutely nobody worries might be insecure while all of their users are getting QR code…

Jason Kikta

Yeah, absolutely. I'm thinking about third SaaS vendor from the left and, you know, small kind of niche open source project being able to do formal verification on the crypto libraries that they're using in their implementation of it. And I like that future, I think, is amazing. And so, you know, I…

Patrick Gray

Yeah. I mean, don't forget, too, how complicated this sort of formal verification is. I mean, for a long time, everybody would talk about the SEL4 Linux kernel or the microkernel because it was basically the only thing on the planet that had been formally verified. Australian project, mind you. So…

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.
“There's a human in the loop, but that human still has to determine the objective. They have to select the targets. They have to discern the value of the exfiltration, because the model's just going to blast statistics at you that make it sound really useful when maybe it isn't. This is faster, but it's not a bunch of new techniques that defeat existing security methodologies. It becomes way, way less forgiving of an environment for gaps and errors by the security team.” — Jason Kikta, Risky Business

Patrick asked whether Google's report on agentic hacking crossing a threshold was pitched about right. Jason says the research is accurate but that these are automated workflows, not autonomous intrusions.

Risky Business · 2026-11-18 Permalink → Listen →
Risky Business Around 18:47 into the episode
Jason Kikta

…ways that it got portrayed, you know, it's what that old thing about, you know, the reporter writes the story, the editor writes the headline. And so, you know, I think there's some headline stuff on it that is probably a little bit more exciting than it needs to be. But the glitter cannons, the

Patrick Gray

dancers, the glitter cannons and dancers.

Jason Kikta

And I think this is, but I think it's like, it's really good research. And I think they capture accurately that this is autonomous workflows, right? Essentially, a really enhanced audience. Automation, these aren't antonymous intrusions, right? And so, you know, there's a human in the loop, but that human still has to determine the objective. They have to select the targets. They got to discern the value of the exfiltration because the model's just going to tell you, like, oh, I've got so many data points and it'll just blast statistics at you. That makes it sound really useful when maybe it isn't. And, you know, ultimately at the end, the human has to assess the end state. So, you know, this is faster, but it's not a bunch of new techniques that defeat existing security methodologies, but it becomes way, way less forgiving of an environment for gaps and errors by the security team.

Patrick Gray

100%. And it's funny because in this week's sponsor interview, Ed starts off by saying exactly the same thing, which is the vulnerabilities that are being exploited here are not new. Like nothing is new here. But I think we've got to be careful not to poo-poo it so much because, oh, it's not new.…

James Wilson

Yes, it was the first time using AI in months, if not. Maybe it's actually the first time it felt this way, but I couldn't leave my desk. Once I started to see what it did after the first sort of, you know, enumerate the API endpoints, yeah, that's fine. And then it was poking around on a few pages…

Patrick Gray

I mean, you're going to start with the easiest stuff. Like to me, this makes sense. You're sitting there going, how weird. And I'm like, yeah, start with the easy win, right? You go through, you bypass the capture. Okay, let's just start with the passwords. And that's what I mean about how they…

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.
“I think where it's headed is an increase in parallelization and an increase in stealth. Right now they're not being employed in a massively parallel way. That moves in multiple dimensions: you can be trying multiple techniques in parallel, you can be doing many more intrusions in parallel. This is going to scale up, and it's going to scale up rather quickly. So while we have some time to prepare, now is the time to start reshaping your defenses and your processes, because it's going to go fast and it's going to be ugly.” — Jason Kikta, Risky Business

Patrick asked where offensive AI frameworks and harnesses are heading after discussing the Google report on agentic threat activity. Jason, the new CISO at an AI-safety nonprofit, predicts more stealth and many attacks run at once.

Risky Business · 2026-11-18 Permalink → Listen →
Risky Business Around 22:25 into the episode
James Wilson

…the credentials out? Or was there like, you know, a 48, 72 hour 100 million token effort to extract this? That's the thing I would really like to know is just how, you know, how many steps did it have to go through to get to these credentials is the bit that's kind of missing from these stories.

Patrick Gray

…safety and whatever, but this is like really about harness engineering because like the model, it doesn't really matter what model you're using with these things. But what are your thoughts on where all of this is going with these types of harnesses slash agents like Artex and CAN and Antifox?

Jason Kikta

So I think that where it's headed next is, you know, an increase in parallelization and increase in stealth, right? And so absolutely makes sense that the first thing the model would do is start password spraying because why not? But the problem with that is it's very, very noisy. And so, you know, I think what we're going to see is we're going to see, you know, it evolve really in a few different ways of one of, you know, James Wilson sitting there watching it and the thing blows 24 hours on solving a CAPTCHA when it could have just said, hey, James Wilson hit a CAPTCHA, like solve the CAPTCHA. Human do a human thing. You know, and you see this out of models all the time where they're like, oh, well, I couldn't access this website because it was denied. So rather than asking you to do it, you know, I searched the web for 50 other things and blew through, you know, 10,000 more tokens instead of just saying like, hey, can you like download this into a PDF and attach it? So I think there'll be improvements there. And I think that there will be improvements in, you know, essentially focusing on the stealthy methods first and only getting noisy when you really have to. And then the other thing, and I think we're going to, you know, see this later on in the show as we move on in the news, is, you know, right…

Patrick Gray

Yeah. Speaking of the White House, it's super intelligence task force, its SI task force. It's asking regulators and whatnot in the US to go a bit easy on companies that have been owned as a result of like autonomous hacks because everyone's getting owned basically and sort of singling out and…

Jason Kikta

Yeah. And that's the thing is that, you know, I think right now saying, hey, we don't like, we're a little off the map. Lot of our best practices or some of our best practices aren't holding anymore. And so we need time to adjust. And what was reasonable and prudent six months ago isn't now. That's…

Patrick Gray

Now, last week, I said that people should grab Artex, run it against their own stuff, to which James and Chris Krebs both said, whoa, Pat, maybe check with your lawyers first because these things can go off the rails. I actually think the White House should similarly indicate that unless you've…

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.
“It's long-tail neglect of maintenance, neglect of talent management, neglect of timely upgrades. IT has been the bill-payer for everything in technology for, I would say, probably 25 years. Here comes cloud, here comes remote, name your thing. Now it's AI, and it's one more thing for IT to be the bill-payer for. And guess what? You're losing good qualified people, you're deferring maintenance, and things break.” — Jason Kikta, Risky Business

Reporters were asking security people whether the string of outages was a cyber incident. Jason's answer is that it is the same old underfunding, now with AI added to the load.

Risky Business · 2026-11-18 Permalink → Listen →
Risky Business Around 15:09 into the episode
Jason Kikta

Same as it ever was.

Patrick Gray

…like, we would sit here and say no. But, you know, I guess my question to you would be, what explains all of this? Is it a coincidence or is it more potential for stuff to go wrong given the pace of like updates that everybody's having to do and what? Like, what's going on in your mind here?

Jason Kikta

I mean, to me, it's, it's pretty, you know, I got to wake up to that as well, even though I'm out on the West Coast this week onboarding for the new job. I still woke up to a bunch of reporters pinging me, going, oh, my God, what's going on? What's wrong with NSANet? And, you know, the simple answer is all of this is the same as, again, same as it ever was. It's long tail, you know, neglect of maintenance, neglect of talent management, neglect of, you know, timely upgrades. And, you know, IT has been the billpayer for everything, everything in technology, you know, since I would say, you know, probably for 25 years, right? Like we went through this big spurt of adoption in the 90s. And then like, here comes cloud, here comes remote, here comes, you know, name your thing. And now it's AI and it's one more thing for IT to be the billpayer for. And guess what? Like you're losing good qualified people, you're deferring maintenance and things break, right? Like, I don't think any, you know, anyone who's ever worked at Fort Meade like blinked an eye when they saw that NSA net was down. They're like, I mean, it's goes down all the time.

Patrick Gray

It's funny what you say because you talk to people who work at some of these agencies, right? SIGINT agencies, which are ostensibly like, you know, computer experts and stuff. And then you find out about like, just, you know, I've had little glimpses over the years from some of the, like, some of…

Jason Kikta

system. It's not the AI. It's not the sexy thing. And so it doesn't get just like parking garages on that stupid campus. They don't ever get funded. And, you know, you got to walk 20 blocks to get from your car to work, right? Because it's just, it's not the interesting thing. And so, you know…

Patrick Gray

I mean, it's, yeah, I mean, what is it? A talk I saw from an Arbor Networks guy in 2012. He's who said, frankly, I'm amazed the internet works at all. It just sort of has, yeah, those vibes at the moment. But let's talk also, let's move on to talking about a report that the Google Threat…

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.
“Whether it's Microsoft or the FAA or NSA, it's all the symptoms of the same disease: they're letting go the people who know the dark magics about how the network works. And they're trying to replace it all with a bunch of hastily jammed-in AI that is improving in capability every day, but that doesn't mean it's ready to replace wholesale sets of human beings. And this is just going to keep happening.” — Jason Kikta, Risky Business

Patrick Gray asked whether the recent wave of outages, including air traffic control and an NSA network failure, is coincidence or a sign of rushed change. Jason Kikta says it is long-running neglect of maintenance and talent, with AI now the latest cost IT has to absorb.

Risky Business · 2026-11-18 Permalink → Listen →
Risky Business Around 16:22 into the episode
Jason Kikta

…for. And guess what? Like you're losing good qualified people, you're deferring maintenance and things break, right? Like, I don't think any, you know, anyone who's ever worked at Fort Meade like blinked an eye when they saw that NSA net was down. They're like, I mean, it's goes down all the time.

Patrick Gray

…and stuff. And then you find out about like, just, you know, I've had little glimpses over the years from some of the, like, some of the stuff that happens on those networks, like the operating systems and stuff that they're, that they're running, which are like ancient. Because it's not the SIGINT

Jason Kikta

system. It's not the AI. It's not the sexy thing. And so it doesn't get just like parking garages on that stupid campus. They don't ever get funded. And, you know, you got to walk 20 blocks to get from your car to work, right? Because it's just, it's not the interesting thing. And so, you know, whether it's, you know, Microsoft or the FAA or NSA, it's all the symptoms of the same disease of like they're letting go the people who know the dark magics about how the network works. And they're trying to replace it all with a bunch of hastily jammed in, you know, AI that is improving in capability every day, but it doesn't mean that it's, you know, it's ready to just, you know, replace wholesale sets of human beings. And this is just going to keep happening. And I think a little bit of this too is, you know, especially because of the FAA, like that was the really big one. Like everyone is looking now at every single outage, like, oh my goodness, here it is. It's the big one. And, you know, the internet's in a constant state of breaking. So it's no surprise.

Patrick Gray

I mean, it's, yeah, I mean, what is it? A talk I saw from an Arbor Networks guy in 2012. He's who said, frankly, I'm amazed the internet works at all. It just sort of has, yeah, those vibes at the moment. But let's talk also, let's move on to talking about a report that the Google Threat…

Jason Kikta

I think the content is right on. You know, some of the ways that it got portrayed, you know, it's what that old thing about, you know, the reporter writes the story, the editor writes the headline. And so, you know, I think there's some headline stuff on it that is probably a little bit more…

Patrick Gray

dancers, the glitter cannons and dancers.

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.

Collections They Appear In