From our other sites
The story
Behind the recent surge in unauthorized access is a new reality: AI now handles reconnaissance, intrusion, and data theft automatically, and high-performance Chinese-made LLMs (large language models) are circulating for free. An article reporting on this became a hot topic on 5ch. In the thread, many pointed out that high-performance open-source LLMs that run even on low-spec PCs are being released for free, lowering the barrier to learning cyberattack techniques. At the same time, pushback emerged against the simplistic framing of “China is the mastermind,” with some arguing that “they’re just using another country’s LLM as a stepping stone — the people doing it aren’t necessarily Chinese.” The gap between the technical reality and the oversimplified interpretation became the central point of debate.
Behind the surge in unauthorized access, the rise of “Chinese-made LLMs” — AI automatically handles reconnaissance → intrusion → data theft. “It’s rarer to find someone whose info hasn’t leaked” — experts sound the alarm. What can be done?
Sun, Oct 18, 17:00
TV Kochi
Source: news.yahoo.co.jp / Original article here
What people said
By the way, all the big-name American LLMs have that turned on by default.
Ugh… I want to comment but I'll hold back lol
You can easily route around that though
Why isn't anyone doing that?
Because nobody takes any responsibility even when data leaks from a breach
Which basically means the CCP doesn't care either lol
Nah, they've got tight restrictions on that, don't they?
Otherwise the government would shut it down in an instant
Exactly
Cutting the line with China isn't that simple
You'd have to sever China's connections with every other country too — that's the only way it'd work
But wouldn't that choke the bandwidth?
Japanese LLMs don't even come up in the conversation
Pretty sad, honestly
Well, Japan doesn't do distillation using American LLMs, does it
Though, well, a good chunk of hacker groups probably are Chinese anyway
The real issue is China's own lack of literacy — using other countries' LLMs as a stepping stone
What kind of restriction are you even talking about?
Banning open source?
Open source?
Is the LLM in this article actually about local, offline setups?
misunderstandings are going to spread again, weird regulations will get slapped on, and Japan will end up falling behind all on its own.
> What a "local LLM" is
No internet connection. I think it's fair to understand it as AI that can be fully downloaded and run locally.
> Who's making and releasing them
All kinds of players release them — even Google puts some out. The high-scoring ones that run on low-spec hardware tend to be Chinese-made.
> What the actual problem is
The learning cost for people doing the hacking has dropped. The real problem is that hacking tools can now be built far more easily.
Wouldn't you need an enterprise-grade GPU server to even pull that off?
Lately main memory can be used much more efficiently, so fairly smart large models run fine even on an ordinary PC
I think it has plenty of performance.
Being able to monopolize the resources is especially powerful, I'd say.
It's free because it's a local AI, lol
"Cheap" doesn't even cover it — it's handed out for free
All that's left is training it yourself for hacking purposes
Why do you think that is
Because everyone's looking out for their own skin
"Stop reusing passwords,"
"Use multi-factor authentication,"
and "Watch out for suspicious emails" —
the Digital Minister just stated three perfectly correct countermeasures and still got bashed for it. Poor guy.
Lol, those are literally the first three things an AI tells you if you ask it
No matter how much individuals protect themselves, it's pointless if companies keep leaking the data anyway
At that point you might as well say "just don't register your personal info anywhere"
I don't think it's doing anything that sophisticated.
It's just brute-forcing its way to finding vulnerabilities and then exploiting them.
So it's not like it can hack the Department of Defense's computers.
It's not hacking into a bank's core accounting systems either.
It's just finding the vulnerable company among the millions out there.
Background and key points of this discussion
What this report points to is the fact that open-weight large language models released for free by Chinese companies can be downloaded and used by anyone. As the thread pointed out, the nationality of the model’s creators and the nationality of the person using it to attack are separate matters — jumping straight from “Chinese-made LLM” to “this is the Chinese government’s doing” risks being a hasty conclusion. That said, the fact that AI can now take over brute-force vulnerability hunting and the building of hacking tools — meaning even low-skill attackers can target huge numbers of companies — should be taken seriously as a real threat, just as the experts warn. The countermeasures the Digital Minister cited in the article, like “stop reusing passwords,” are correct at the individual level, but as the thread also noted, much of the data leakage stems from vulnerabilities on the corporate side — leaving a structural problem that individual precautions alone can’t fully solve.
*This article is excerpted and summarized from the 5ch (News Express+) thread “[Unauthorized Access] Behind the surge, the rise of “Chinese-made LLMs” — AI automatically handles reconnaissance → intrusion → data theft. “It’s rarer to find someone whose info hasn’t leaked”.”
Leave a Reply