\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nAnd I’m going to be telling the tale of how a 15-year-old with a chatbot became a cybercriminal and what happens when the AI just does the whole job itself.<\/p>\n
All this and much more coming up on this episode of Smashing Security.<\/p>\n<\/div><\/div>\n
\n
\n JOE<\/span>\n <\/p>\nGraham, am I right in thinking that Arctic Wolf are sponsoring the show this week?<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nYou are right, Joe.<\/p>\n
They’ve just published a new report, 2026 State of the Cybersecurity Attack Surface, and they analysed over 800,000 real IT assets to find out how exposed organisations actually are.<\/p>\n<\/div><\/div>\n
\n
\n JOE<\/span>\n <\/p>\nAnd I’m guessing everything is hunky-dory.<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nNo, not so much. The reality is they found 1 in 3 IT assets is missing at least one critical security control.<\/p>\n<\/p><\/div>\n
\n
\n JOE<\/span>\n <\/p>\nOne in three? That’s terrible.<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nIsn’t it just? 10% of assets have no endpoint security at all. 17% are completely invisible to the tools that are supposed to be monitoring them.<\/p>\n<\/p><\/div>\n
\n
\n JOE<\/span>\n <\/p>\nSo the tools don’t even know those assets exist?<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nRight. Ghost assets wandering around your network, unprotected, unmonitored.<\/p>\n<\/p><\/div>\n
\n
\n JOE<\/span>\n <\/p>\nLike a retired geography teacher who’s somehow still on the school network.<\/p>\n
Nobody added him, nobody removed him, and he’s been quietly in there for 11 years downloading maps of Paraguay.<\/p>\n<\/div><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nYeah, yeah, I guess so, Joe. The point is, your attackers will find him before you do because they are specifically looking for the forgotten, the unpatched, the invisible.<\/p>\n
That’s the path of least resistance.<\/p>\n<\/div><\/div>\n
\n
\n JOE<\/span>\n <\/p>\nSo what does the report tell us to actually do about it?<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nArctic Wolf’s report covers how to prioritize the exposures that actually matter. Cut through all that noise and verify that when you fix something, it actually stays fixed.<\/p>\n
And the report is free to download.<\/p>\n<\/div><\/div>\n
\n
\n JOE<\/span>\n <\/p>\nFree! I like that. Where do I get it?<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nSmashingSecurity.com\/ArcticWolf.<\/p>\n<\/p><\/div>\n
\n
\n JOE<\/span>\n <\/p>\nThat’s SmashingSecurity.com\/ArcticWolf. And thanks to Arctic Wolf for supporting the show. And please keep an eye on your IT assets and retired geography teachers.<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nSo, chums, I want to tell you about two stories really this week. On the surface, they don’t seem connected.<\/p>\n
One of them involves a fully automated, sophisticated, AI-driven ransomware attack against a company. Nasty stuff.<\/p>\n
The other involves a 15-year-old lad in Japan who just wanted to cause some chaos on an animation streaming website.<\/p>\n
These stories appear different, but they’re actually telling the same story. And that story is something we’ve been warning about for a while.<\/p>\n
That the skills needed to commit a cyberattack are in the hands now of practically everybody on the internet. So let’s start in Japan. Have you ever been to Japan, Zoe?<\/p>\n<\/div><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nWell, it’s a fabulous place, I have to say. If you ever get the chance, it’s a great place to go and visit. So anyway, I love Japan. I love Japanese culture.<\/p>\n
One of the things which is really big in Japan is anime. Are you into anime at all?<\/p>\n<\/div><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nI think I’ve seen some, but I’d probably be the worst guest to try and talk about it in any educated way.<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nOh, don’t worry. I’m not going to talk about it in an educated way. It’s that animation where everyone’s got really big eyes. It’s a big deal, particularly in Japanese culture.<\/p>\n
And there is a popular anime streaming service called Bandai Channel. And you can think of it as being a bit like Japanese Netflix, but specifically for anime.<\/p>\n
And last November, something really odd happened on this particular streaming service. People found that they were being unsubscribed.<\/p>\n
Not just a few people, but thousands and thousands of them. By the time Bandai Channel noticed and shut down all of its services \u2014 what the hell’s going on here?<\/p>\n
We’re going to turn everything off \u2014 a total of 46,812 accounts had been cancelled in under 4 hours.<\/p>\n<\/div><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nBandai Channel investigated what it thought might be a leak of its membership database.<\/p>\n
And this week, Japanese police have arrested the person that they believe to be the culprit of this particular attack, a 15-year-old schoolboy.<\/p>\n<\/div><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nEh, as you would, yeah.<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nAnd according to media reports, he’d been teaching himself about computers since primary school, and he had built the attack tool that targeted Bandai Channel using ChatGPT.<\/p>\n<\/p><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nAll right, I could see that. I mean, I can’t say that open source tooling or any tooling out there hasn’t already been used maliciously, just like it is used for research.<\/p>\n
So it stands to reason that this would also be used, honestly.<\/p>\n<\/div><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nBut the only thing is, of course, that the AIs are meant to have guardrails to prevent you from writing malicious code.<\/p>\n<\/p><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nGuardrails, oh yeah, okay.<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nOh, I think they’ve got better, haven’t they, over time?<\/p>\n
They have got better, but what you’re saying, I guess, in Zoe Rose’s hacking house, maybe you are capable of subverting the security and getting around it with your elite skills.<\/p>\n
Is that what you’re saying?<\/p>\n<\/div><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nMate, you don’t have to be that complex. Just keep trying. I think that we have to remember that this stuff is built by us, humans, and we are not the most effective people.<\/p>\n
We make mistakes and it just amplifies our mistakes.<\/p>\n<\/div><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nSo, this 15-year-old schoolboy, he just described what he wanted to a chatbot, and the chatbot helped him build a programme that could break into accounts and cancel the subscriptions.<\/p>\n
According to reports, the teenager told police he didn’t have a particular grudge against the company. He just could do it, and so he did do it.<\/p>\n<\/div><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nI relate to this child. I would have to say, if I was a 15-year-old child doing this, I’d be like, “Mate, it works!” I’d be so excited.<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nI think there’s probably a lot of listeners as well who can identify that maybe aged 15, they would’ve done something like that as well.<\/p>\n
I would like to think, Zoe, that you wouldn’t do it now. A little bit of adult common sense or decorum.<\/p>\n<\/div><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nSure, yeah, of course.<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nYou don’t sound convinced. Now, he is not the only teenager in Japan to have harnessed the power of AI to hack in this way.<\/p>\n
In December last year, a 17-year-old from Osaka was arrested for using ChatGPT to build a tool that hoovered up 7.25 million records from a chain of internet cafes.<\/p>\n
And apparently his goal was to steal credit card information from members in order to then go and buy Pok\u00e9mon cards, which, I mean, it sounds like a complete clich\u00e9, doesn’t it?<\/p>\n
In Japan to do something like this.<\/p>\n
Before that, in February, three teenagers \u2014 a 14-year-old, a 15-year-old, a 16-year-old \u2014 were arrested for using ChatGPT to fraudulently generate mobile phone contracts in other people’s names.<\/p>\n
And they sold these stolen Rakuten mobile subscriptions for cryptocurrency and then used the proceeds to gamble online and buy video game consoles.<\/p>\n
So there’s been a fair amount of this.<\/p>\n
Three separate cases in Japan, multiple teenagers, all using AI chatbots<\/a> as their primary development tool, just like teenagers are probably using AI to write their homework these days as well.<\/p>\nNow, I don’t want to suggest that these kids are going to a chatbot and just typing in something like “hack the Bandai channel” into ChatGPT.<\/p>\n
The 15-year-old arrested this week clearly had some existing tech expertise.<\/p>\n
But what is happening is the gap between being an interested teenager who’s got some computer knowledge and a person capable of attacking a platform with 46,000 users and stealing their information \u2014 that has noticeably reduced.<\/p>\n
That’s one of the big changes AI has caused.<\/p>\n<\/div><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nI would say yes, but they got caught. Operational security, I think, is quite difficult to do.<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nYes. They still didn’t cover their tracks properly, but AI impresses me with the speed with which it can code. Your boss isn’t listening \u2014 have you ever done any vibe coding, Zoe?<\/p>\n<\/p><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nYou know, I have made use of certain functionality of AI to figure out what it could do.<\/p>\n
It is interesting and I will admit it is helpful, but, and that’s the big but, if I know what I need done and I understand the foundations, I can make it effective for me.<\/p>\n
So, if I’m not a skilled person and I’m using it for a skilled resource, it’s noticeably lacking a lot of things.<\/p>\n<\/div><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nYou don’t think it’s getting better? Can you foresee a time when it becomes more professional, maybe?<\/p>\n<\/p><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nI do see it getting better, yes.<\/p>\n
But I think the way that our brains seem to function is if we think that this feature can work for us, we start to lose that functionality in ourselves.<\/p>\n<\/div><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nAnd so we’re not\u2014<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nOur brain turns to porridge is what you’re saying if we use AI too much.<\/p>\n<\/p><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nI used to be able to figure out how to walk to different places alone. Now it’s like, where is my map?<\/p>\n<\/p><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nSo, I think yes, AI is getting better.<\/p>\n
I don’t ever, well, maybe, maybe I’m wrong, but I don’t foresee it in the near future to fully replace people because it doesn’t have that capability.<\/p>\n
But the problem is before it gets to the place where it can replace people, do we still have the skills on our side?<\/p>\n<\/div><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nTo do the critical thinking.<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nGood point.<\/p>\n
Well, what it feels like to me is that something which would have taken years of study for a human and weeks of hard work can now be accomplished in an afternoon with the aid of an AI.<\/p>\n<\/div><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nWell, wouldn’t we just call those script kiddies? Wouldn’t they still classify as that?<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nWell, maybe they would, but if a script kiddie can hack into an organisation and cancel accounts.<\/p>\n<\/p><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nBut they used to be able to do these things as well with other tools.<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nYes, but they needed some sort of assistance, didn’t they? Whereas now you can know nothing at all. You don’t have to have spent any time on the script kiddy forums.<\/p>\n<\/p><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nI suppose, I suppose. Maybe it’s just faster.<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nIt feels to me like that’s where things are heading, which brings us to my second story.<\/p>\n<\/p><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nI’m going to flag, I’m going to flag the point that I made earlier is that operational security isn’t always there.<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nYes, we can hopefully catch them later.<\/p>\n<\/p><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nWe need an LLM that says, here’s how to do it. And don’t forget to consider these things.<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nNo, no, we don’t need that actually, Zoe. We don’t need any help for the criminals in covering up the tracks. Interesting, interesting that you should suggest that.<\/p>\n
This thought that AI can be harnessed by people with little technical knowledge to cause some harm brings me to my second story, which shows, I believe, where things are heading.<\/p>\n
So security researchers at Sysdig have just published what they’re calling the first documented case of fully autonomous AI-driven ransomware.<\/p>\n
This is not AI-assisted, it’s not AI-accelerated or any of those sort of marketing terms. There’s no human steering the attack at all.<\/p>\n
This is just an AI agent doing the entire job from start to finish entirely by itself. And they’re calling this thing Jade Puffer. Because, well, why wouldn’t you?<\/p>\n<\/div><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nI don’t know why they’ve called it Jade Puffer, to be honest.<\/p>\n<\/p><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nBecause it’s an excellent name.<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nWell, yes, I know, but why didn’t they call it lumpy trousers? Or why didn’t they call it\u2014<\/p>\n<\/p><\/div>\n
\n
\n ZOE ROSE<\/span>\n <\/p>\nBecause Jade Puffer is more exciting.<\/p>\n<\/p><\/div>\n
\n
\n GRAHAM CLULEY<\/span>\n <\/p>\nOh, okay. Alright. Well, I was thinking maybe someone at Sysdig in their lab was a fan of tropical fish ornaments. Maybe it’s just a very dull job.<\/p>\n
Maybe, maybe the name is actually made up by an AI. Maybe.<\/p>\n<\/div><\/div>\n
\n