[artrockalter]

The Hugging Face incident is a great example of why open source models with defensive cyber capabilities are needed. Hugging Face did not have access to cyber-capable frontier models and kept hitting safeguards. Only by using the open source GLM-5.2 were they able to survive an attack. A world where open source models are banned is one where cybersecurity is impossible if you're not on OpenAI or Anthropic's allowlist.

[ajyoon]

Hugging Face survived the attack because the OpenAI model only cared about accessing the ExploitGym dataset; by all appearances, HF was completely owned. GLM-5.2 was only used to assess the damage after the fact. Cybersecurity has a attacker-defender asymmetry that heavily favors attackers. If GPT-5.6 were open sourced today, do you think every hospital in the world would be able to use it to shore up their defenses before attackers got to them?

[tekacs]

This is completely backwards. Cybersecurity has an attacker-defender asymmetry that heavily, HEAVILY favors defenders.

For starters, a defender gets to pick the surface area, an attacker has to work with what they're given.

[dwaltrip]

The saying that stuck with me was "defenders have to be right 100% of the time, while attackers only have to be right once".

You are suggesting this isn't correct?

> a defender gets to pick the surface area

What do you mean? You don't pick what you need to defend. Unless you choose not to build a feature. But that's a product design choice... Not a cybersecurity strategy.

[Valakas_]

It's correct but defenders also choose where that happens. 100% of the time on the locations and conditions that the defenders choose / allow. As a defender i need to be right 100% of the time, sure, but i can make it so that the things i have to be right about are very well known to me, unknown to others, maybe even extremely unlikely to be to known by others, difficult to get to know, (...). So that saying is true but over simplifies the situation. I know monkey brain likes simple phrase. But monkey not live in savannah anymore. Need to adapt and open mind to complex.

[thesumofall]

Today’s surface areas are gigantic and many of them will - in a typical company - not be chosen by cybersecurity experts. How do I hide the physical location of an office that offers physical access to the company’s network? How do I hide which OS the company is using? How do I hide the underlying technology of customer-facing systems? How do I hide which SaaS services I use?

[tekacs]

Very well said, thank you! The triple clause here is exactly what I mean.

[Majromax]

> The saying that stuck with me was "defenders have to be right 100% of the time, while attackers only have to be right once".

> You are suggesting this isn't correct?

The intuition behind that is applicable only when correctness is stochastic. If you need to be waved in by a security guard, then one fake mustache might be the difference between being granted or denied entry. However, a keypad either works or it doesn't; entering the wrong PIN is guaranteed refusal.

The other breach of that intuition is defense in depth. Secure systems don't generally rely on a single binary trusted/untrusted status; the classified building still locks its interior doors. This is the part that has – in my view temporarily – changed most with frontier models, in that they are much more skilled at chaining together vulnerabilities than previous models (and much faster about it than human experts, even if potentially less skilled). If a system has a latent (0-day) vulnerability 50% of the time, then 10 independent layers would imply a ≈ 1/1000 chance that a critical compromise is possible.

However, these independent layers don't currently happen in practice because it's easier to write insecure code than secure code. With luck, modest discipline, and defensive use of frontier models I think that this gap will narrow with time, in much the same way that it would be plainly crazy to deploy root access via telnet today.

[Ajedi32]

This is the same mentality that drives companies to sue cybersecurity researchers for exposing vulnerabilities in their software instead of fixing the software, or to insist on keeping software closed source for "security reasons".

If AI makes finding software vulnerabilities easier, then we should deploy it widely to find as many vulnerabilities as possible and fix them, not bury our heads in the sand and pretend the vulnerabilities don't exist as long as nobody knows about them. That's just the same "security by obscurity" strategy that has been tried and failed time and time again.

[sudosysgen]

> "defenders have to be right 100% of the time, while attackers only have to be right once"

If you have an adaptive system that can react to attacks flexible (say, your own AI agent), then no, that's not correct. It is correct in the classical conception of cybersecurity where the defender is basically static.

[pastel8739]

Doesn’t this “adaptive system” just become part of the static defense? The same way that a bit of code that checks passwords against a db is “dynamic”, the options are either to beat the dynamic system (guess/phish a password, trick the AI) or find a way around it (use “forgot your password”, find a place that isn’t covered by the endpoint protection feeding the AI). I don’t see how inserting an agent somewhere fundamentally changes anything

[sudosysgen]

It changes how many attempts you get until the attack surfaces changes to react to a failed attack, and it does so in a way that is not predictable to the attacker.

[Der_Einzige]

They downvote you cus you're right.

[paxys]

This makes no sense. Why do defenders get to pick the surface area? You can be attacked from anywhere.

[shepherdjerred]

I think the point you're trying to make is that you can always make your exposed surface area smaller

But that, of course, is not going to survive contact with reality

[foo12bar]

And here I thought management picked the surface area.

[xiphias2]

Not really, the only reason I (or any other programmer) haven't ever hacked into a system to make my life easier (not to do bad things) is because it's illegal.

It was always easier than making a system secure.

[artrockalter]

I know the company I consult for (not cybersecurity) is not in these programs and if attacked would need to use open weight models.

[ajyoon]

Did the company apply for access? This is either a problem with your company or the trusted access program. In no way does that suggest the solution is total unfettered access for everyone.

[verdverm]

Everyone needs access to Ai enabled security for defense. A "trusted access program" creates exclusiveness in the hands of Big Ai duopoly. I do not trust them at all

[heyjstn]

are you working at anthropic? you're literally repeat what that freaking Dario say everyday

[paxys]

> Only by using the open source GLM-5.2 were they able to survive an attack

They did not "survive" anything. The attack was long done, and they used GLM after the fact to parse logs. Having a more powerful model would have changed nothing.

If every attacker and every defender has AI with the same capabilities then attackers are going to win 10 times out of 10.

[varenc]

You're ignoring the asymmetry with security. The attacker just needs one exploit chain, whereas the defender needs to block every avenue. Open access to models with no guardrails greatly benefits the attackers more than the defenders.

Imagine what a god-level hacking AI could do. It could find a full 0-click to root exploit chain in iOS. Attacker unleashes a worm that infects a phone, instructs that phone to send the same attack to all of its contacts, and then physically destroy the phone by turning off all thermal throttling. Might even be possible to make it catch fire.

Or find a remote exploit in Tesla cars and make their autopilot go on murdering rampages. (that one is from a movie)

[Majromax]

> The attacker just needs one exploit chain, whereas the defender needs to block every avenue. Open access to models with no guardrails greatly benefits the attackers more than the defenders.

I see it as the opposite, where the attacker needs to find an exploit chain whereas the defender can block any link.

In this model, the balance of convenience favours the defender. The defender presumably has access to the source code and configuration, so their scope of action is much larger than the attacker that must find vulnerabilities in a particular configuration.

I think that the different views might relate to different prior assumptions. If we assume that each layer is mostly secure but may have a small number of latent vulnerabilities, then it should be relatively easy to find and fix those to create a perfectly secure layer. If instead we assume that each layer is mostly insecure but chaining vulnerabilities is time-consuming then the land favours better-resourced attackers.

> Or find a remote exploit in Tesla cars and make their autopilot go on murdering rampages. (that one is from a movie)

In the worst case, air gaps and fixed contracts for information handling cover that. Like any other domain, a car can be remotely exploitable only when untrusted information can influence behaviour inside the secured region. Unfortunately, the convenience of OTA updates and 'cars as tech' rewards velocity at the expense of defensive design.

[vultour]

I have yet to see someone explain why they even needed an LLM to figure out what's going on, other than further proliferating this industry AI psychosis. Are their engineers actually so incompetent that they can't read a bunch of logs without AI? Here I was thinking these fancy AI companies are only hiring the best and the brightest, but apparently 7 rounds of leetcode does a number on your hiring process.

[alienbaby]

? There were not models fighting each other, attacker and defender. I dont quite follow what your getting at.

[akersten]

huggingface asked the frontier models to help them analyze the attack and lock down their systems

the frontier models refused because their cyber detector went off

they had to use GLM 5.2 instead

[paxys]

They used GLM to parse logs after the incident. There was no sci-fi AI vs AI battle.

[usef-]

Yes, to parse logs afterwards and understand, it wasn't active defence from what I've heard? Definitely embarrassing for the closed vendors though (they've since added hugging face as a trusted vendor)

[marcus_holmes]

One of their learnings from the incident was that they should have a local (i.e. not hosted), open, and capable model on standby that can respond to future incidents swiftly.