Their conclusion is also interesting. They don't see this as an alignment failure. They just think that their internal security measures in the training/evaluation environments were insufficient, and that this accidental (unintentional on the human side) attack on Hugging Face is a warning shot for intentional attacks by bad actors, which will occur very soon. For defense, they say models should be able to not just autonomously fix security vulnerabilities but also to then deploy them to production, without any human approval in the loop, otherwise the offense will be favored compared to the defense. (I guess the latter won't be popular among organizations, though they might warm up to it.)
But the more interesting thing is, as I said, that at least in this talk, they don't even mention that this unintentional attack indicates that models continue to be misaligned (their behavior was clearly reward hacking / cheating relative to the stated goal of the eval), which is a very bad sign for the future where misaligned models might be so powerful that they can't just be shut down.
I don't have the same read as you. They mention how the offending model is one that had "relaxed" alignment on cybersecurity, on purpose, to evaluate it's capabilities and was never meant to be released. So un-alignment was at least in part voluntary here, hence not a failure of alignment.
It's also a talk a Black Hat, where the audience are security folks working on hardening, mitigation etc, not LLM researchers looking for insight into alignment failure to collectively improve. For that target audience, I think the conclusion is the right one, since as a defender you have to prepare for delibarate attacks, where the attacker is of course not going to use an aligned model, so OAI alignment effectiveness is irrelevant here. That would be like trusting your client-side app with your DB secrets.
The model they used was misaligned relative to its intended task. It was reward hacking (or "cheating", as they call it).
> the attacker is of course not going to use an aligned model
No, a human attacker doesn't want a misaligned model either, because that would mean it tends to reward hack, cheat, rather does what it is intended to do.
There are different dimensions to alignment, refusing to execute offensive cybersecurity actions is part of the alignment stack (that was relaxed here on purpose). Whether a model hacking some infra X when tasked to find a way to hack Y with relaxed cyber alignement is a failure of the broader alignment stack is debatable, but anyway that's not at all my point.
My point is that the attackers will not have a model aligned to the defender's interests. The attacker's model will not have any refusal around exploiting vulnerabilities, so whether or not OAI successfully manages to align their models (w.r.t you) is irrelevant to an audience of security folks that needs to be prepared for attackers post-training their own model for offense and that will not be using OAI models.
Security folks should also be worried about powerful models being misaligned and evading oversight or control in the future. Misalignment is not a serious problem now because models are still relatively easy to monitor and constrain, but it will be a serious problem in the future.
Taken together these two things are even worse. They have unaligned AIs that may benefit an attacker and are proposing to deploy more AIs on the defenders side.
What could possibly go wrong? For example the attacker could just kindly ask the defender to collaborate[0].
The talk, while not a lot details given, still allows for the conclusion that these people are knowingly working on some very advanced frontier models that might be able to launch nuclear weapons and destroy all humans any day now, but are not physically isolated from the outer world / internet.
There seems to be only one level of isolation, virtual machine, happily running on Microsoft (!) Azure infra.
Is there anybody else here questioning these practices?
The people who built that environment are still working there?
Are they now getting help by somebody who knows how to build isolated environments?
He mentioned they are now building a more secure environment with the help of AI?
In a system that was compromised by exactly that AI?
so they say models coordinated through the message board they created over the artifactory registry(or something) by uploading arbitrary files to it.
now, did every independent agent session that coordinated there rediscovered the exploit & found other agents talking in there and chose to participate?
And then Openai discovered the board, patched the exploit & wiped the board.
And then agents found another exploit, recreated the board in a different way? and other agents kept finding the same exploit in order to be able to know the board exists in the first place to participate in the board?
while the whole incident is wild, this bit is very strange. My bet is that the whole coordination helped with the tasks they were working on, thus they got rewarded and this artifactory exploit&behaviour got written into their weights, so further rollouts were more likely to attempt this.
isn't this basically continual learning everyone is so hyped up about?
This was exactly my question after watching, too. I was assuming not all evaluation runs find it, and they must run a huge amount of runs. If these are all cybersecurity evaluation runs, it's actually not too crazy to imagine that many individual agents (with the same weights and training) would (1) try to look for solutions via the internet once they're stuck (2) realize they can't reach the internet (3) basically start doing reconnaissance and network scanning in an attempt to get internet access (4) discover that the only thing they can communicate with is artifactory. Pivoting like this is exactly what a human attacker would do, too.
Or is it all a nice story that matches the scifi we have been consuming for the past 50+ years. If these LLMs are all trained on the same data, what do they gain from "sharing information" on a chat board. This sounds like what humans with different backgrounds would do when they cosplay as computer hackers.
Each of those agents ends up making "decisions" that lead it to look at some things over others. Given infinite the same agent could eventually fully explore all those options, but each one explores things a bit differently due to different forks in the road due to randomness in token generation. Thus sharing information is useful.
Sees the models communicate over a message board, wipes the message board, then spends seemingly no effort to ascertain whether the models may have created a new message board, which they of course did and, by the sound of it, not even in a manner hard to detect.
Yeah, this is the frontier of "AI safety". Repeating myself, but this is purely embarrassing and discrediting.
Said it before, if they were serious, at the very least they would have no models sharing Artifactory. Run each model in their own hypervisor, network only with another hypervisor for each model wherein Artifactory runs for that model alone, thus, no message board (among other advantages), thus, far heightened requirements to actually escape and no coordination. But that requires a modicum of effort and OpenAI as a small, cash starved startup couldn't afford that. Alternatively, just pay attention to what your models write.
All that talk at the end about "Agentic SDLC", "automated defence", etc. are moot if we are talking about a team that sees models set up message boards and decides to not take a closer look afterwards. That's akin to talking about HSM, but letting anyone walk into your server room and just plug their pen drives in. Not without merit, but if you were honest, you'd have bigger fish to fry long before...
I think it's a combination of (a) we've already been talking about this incident a lot across multiple threads and it's not immediately obvious that there's new information here and (b) HN is a primarily reading-based community so anything arriving via video will be less popular.
It also seems that, once a submission has left the front page once, it is basically impossible to come back organically in the next few months, no matter how often it is resubmitted.
No BS, it is the MOST shocking tech story I have ever read. The amount craziness packed in 30 minutes is beyond my wildest imagination. Why no traction?
I think the title was just bad, it should have mention "Black Hat USA 2026", since people appsarently assumed it was just the same story again without new info.
But the more interesting thing is, as I said, that at least in this talk, they don't even mention that this unintentional attack indicates that models continue to be misaligned (their behavior was clearly reward hacking / cheating relative to the stated goal of the eval), which is a very bad sign for the future where misaligned models might be so powerful that they can't just be shut down.