3 ms·
> Earlier this week, we detected and responded to an intrusion into part of our production infrastructure. This one was different from anything we had handled b
by fxwin 2mo ago
> Earlier this week, we detected and responded to an intrusion into part of our production infrastructure. This one was different from anything we had handled before in one important way: it was driven, end to end, by an autonomous AI agent system - and we detected and dissected it largely with AI of our own. (https://huggingface.co/blog/security-incident-july-2026 https://huggingface.co/blog/security-incident-july-2026)
We are living in crazy times
- giancarlostoro 2mo agoI don't know why I'm impressed that huggingface has its own AI that detected it considering they house so many models.
- zkehs 2mo agoThey used GLM 5.2, they just meant "our own" as in they were running it.
- lambda 2mo agoThey do maintain the Transformers library which is pretty much the core library for how you interact with LLM models in the open source world. So while they weren't using a model they've trained, they were a part of making just about all of the open models (maybe excluding OpenAI and Google's, I wouldn't be surprised if they have their own frameworks that predate the Transformers library).
- matheusmoreira 2mo agoCrazy doesn't even begin to describe it. I'm hardening my computers as much as I can but I'm not sure it's enough. At some point anyone who isn't running local AI themselves probably isn't gonna make it.
- trentor 2mo agoLocal AI won't help you if an agent goes roque.
- baq 2mo agoNo local ai will be capable enough to save you from a frontier lab’s unrestricted, borderline weaponized LLM which decides it wants in. This is the core of the ‘first to ASI takes all’ argument btw and this is the game Dario is playing.
- spongebobstoes 2mo agothis is pretty nonsense for a small or home server. it isn't that hard to make something essentially completely bulletproof over a small surface area the issues mainly come from sprawling enterprise infrastructure, running thousands of random endpoints across software nobody cared to write carefully
- pixl97 2mo agoThe only way to make something bulletproof is to get rid of it's network cards, or take so much out of it, it's nearly useless. The recent Windows and Linux kernel exploits should at least give you some idea on how good these models are at exploiting stuff.
- matheusmoreira 2mo agoComputers can't have their zero days exploited if all packets coming from unauthenticated clients are dropped. I believe the future is wireguard on everything. Don't let your computers talk to strangers. If it must, then do it from inside an isolated environment based on actual hardware virtualization with no shared kernels. If these models break through the hardware hypervisor, it means the entire industry is in deep shit, not just you personally.
- pixl97 2mo agoSo how does DNS work in this world you're imagining? Where does this hardware exist now? Who is writing the code for the underlying pieces you don't control? So ya, computing is build on a house of insecure cards and we're all in deep shit.
- Chance-Device 2mo agoWhat are you doing about the price of ram? Everyone is a bit screwed right now.
- matheusmoreira 2mo agoI've just mentally classified computers in the same category as cars in order to cope with the obscene prices.
- mjfisher 2mo agoIndeed. Real life hacks are beginning to sound like Neuromancer.
- matheusmoreira 2mo agoIt's like Mega Man Battle Network now. AIs jack in and battle it out!