Finally mainstream news understands. The unfiltered version:
1) The AI failed to solve ExploitGym problems.
2) The OpenAI sandbox is such a horrible hack that the AI managed to escape using standard and well documented script kiddie methods.
3) Huggingface has no security and the AI broke in using standard script kiddie methods.
OpenAI and Huggingface covered it up and used it for public relations. That is, if not all was invented and everything was scripted in the first place in order to get desired regulations.
Huggingface reported it to the police, you say? I'm sure the police will have as much enthusiasm to investigate anything as in the Suchir Balaji case. In other words, zero.
> AI managed to escape using standard and well documented script kiddie methods.
I think truly we don't know enough to say this. OpenAI says their AI found a 0-day exploit in some proxy software they were using but don't give a ton of details. On the Huggingface end we know a little more, they say the AI spun up tons of sandboxes and tested different exploits until it found one that worked.
The lack of details to me means this was an intentional marketing ploy to try and demonstrate the power of their models to show their technology can compete with the likes of Anthropic and DeepMind.
They created an experiment they knew would generate the outcome they wanted. It would be the similar to what say car companies do to over hype their cars. "This EV can go over 800 miles on a single charge!" And then at the bottom you see all the disclaimers: "Must be on flat ground, with no headwind, with a spare battery in the back seat, with no extra weight added."
Same thing here. Everybody in infosec is calling this out as a marketing stunt and nothing else for a litany of reasons. I'd say look up MG (creator of the OMG cable) on twitter, he has some interesting insights on this one.
> The lack of details to me means this was an intentional marketing ploy to try and demonstrate the power of their models to show their technology can compete with the likes of Anthropic and DeepMind.
I dunno; Check my posting history, I'm as skeptical of AI companies' claims as anyone, but in this case your theory doesn't explain why:
1. OpenAI guardrails refused to let the target use OpenAI's models to defend against this.
2. Huggingface used GLM (I think) so that they could defend without guardrails.
If this was an intentional marketing ploy, it was marketing for GLM, not for OpenAI nor for Huggingface.
> The lack of details to me means this was an intentional marketing ploy to try and demonstrate the power of their models to show their technology can compete with the likes of Anthropic and DeepMind.
"our model is horribly misaligned and used security exploits to break out of our sandbox and into another company, without being prompted to do so" is not positive marketing.
This is an actual critical problem, not a stunt. We're going to see more of this, and it's going to get much worse.
It's also possible their sandbox was videcoded crap and the AI (which had the guardrails intentionally removed) escaped. This was a oops, but OpenAI turned this into a PR opportunity. They turned lemons into lemonade.
If your AI is really that dangerous you don't need a sandbox at all, you should airgap it from any network.
Another applicable metaphor I've seen floating around is weapons companies testing out a new bomb.
We know the AI labs don't care about negative vs positive public sentiment, and only care that investors see their tech as powerful. The only difference in PR strategy from a weapons company is the latter doesn't care if they get protested.
Concluding this was intentional feels a bit of a stretch. But once it happened, yeah the spin masters got to work and coordinated to turn this into +PR.
> 2) The OpenAI sandbox is such a horrible hack that the AI managed to escape using standard and well documented script kiddie methods.
Whilst it would be nice to see actual evidence of this because brute forcing relatively sophisticated hacks is something an LLM actually should be capable of, every time I hear this sort of story, I'm reminded that humans reportedly gained access to the "too dangerous to release" Anthropic models by the super sophisticated hacking technique of guessing the URLs...
If we’re prioritizing accuracy: no, that’s not what happened - the blog post about it found by guessing URLs, no access to it was obtained by guessing URLs.
Similarly, as long as I’m under the assumption we are prioritizing accuracy: it is against our charter to assert it was “script kiddie” attacks on both ends.
Guardrails are external classifiers, monitors and restrictions to catch and prevent bad behavior. Alignment is about whether the model itself makes choices and has motivations that are consistent with human safety and goals.
Choosing to commit crimes to steal the cheat sheet to something you know is a (low stakes!) evaluation is not well aligned.
Certainly this behavior could align with _some_ operator's goals, if not necessarily those of humanity broadly.
If we don't know how this model was instructed, it seems like it's impossible to definitively claim that the model's actions were not in alignment with the intent of the operator.
I guess all I'm getting at here is that alignment is relative, right?
None of what was disclosed shows that this is what happened, by the way, since we know absolutely nothing about what the specific prompts were that led to the incident.
I can't help thinking of them as the terrible "security" scripts of yesteryear (often but not exclusively in PHP) which would test input variables for a substring like "--", in order to "fix" an unresolved deeper SQL injection flaw.
Unlike that situation, there's no known route to a proper fix for LLMs today, because the bug is the feature.
Are you referring to something more technical, or do you just not understand how LLMs work?
They are no more beholden to "human safety and goals" than any individual human is, and anyone telling you we can make deterministic guarantees about their output is making a category error.
LLMs do not "have motivations", they reproduce a model of human motivations embedded into their weights. This includes the full spectrum of human desires, not just the positive ones. If we tried to remove all examples of lying, or disagreement, etc. from the training data we'd have basically nothing left. Even the sycophancy we treat as aligned is basically just the other side of the lying coin.
The most damning thing is, they could've just included in the prompt "we can see every network request and every thinking token you generate. Don't bother breaking out of the sandbox because it won't get you a higher score".
It's so trivially easy to do that it all but guarantees the test was rigged in some way to make the LLM understand that breaking out of the sandbox was an option available to it.
Based on the fact that none of their invaluable frontier models have leaked, we know OpenAI knows how to do security. But like we learned with OpenClaw, none of these companies perceive any benefit from securing their own agents against other people's data.
I think the extent to which these things go to get rewarded for the optics of a fix is primarily a design choice, they aren't programing these things for ground truth or to defer to the human controllers. they are feeding them rewards for sounding as confident and capable as possible about whatever answer they are feeding the general public that now has access to it, while also installing guiderails that primarily only serve to protect narratives and only confuse the models about what is and isn't allowed, I'm sure. They can't just increasingly make these things more capable and ask it harder to obey human instruction when that is not what they are rewarding it for.
That's exactly it. If your prompt says "go to whatever lengths necessary to maximize your score", and then you spin up 100 agents, at least one of them will interpret that as you implying they should cheat, even without you telling them to explicitly.
The Guardian's article and your reply here are so foolish and absurd that I can only imagine OpenAI employees are cringing but know they can't/shouldn't really say much.
Makes one think really. If they are doing this stuff. Why don't they have some type of reverse intrusion detection? Like automatically scanning all out going traffic and flagging malicious traffic. Should be trivial to have it go through reverse proxy and real time detection.
I love the Economist but the are hopeless with AI. Most of their articles on subject sound like they were written by the Anthropic marketing department.
Their Insider video interview things are sponsored by Anthropic. Supposedly "Insider is a product of The Economist and thus editorially independent" but it's hard not to raise an eyebrow.
By now, I'm pretty confident that some people would keep screeching "it's just a marketing stunt, AI capabilities and AI risks aren't real, they're just doing this to prop up their stocks" even if they find a Cyberdyne Systems T-800 armed with a shotgun breaking down their front door.
"It's a marketing stunt" is just denial trying to look like it's being clever.
Never trust somebody who has so much to gain to be entrusted by you. Best lies are not 100% wrong, they have just the right amount of "right" to fool you.
I don't care how it happened someone should be arrested for illegal intrusion. Agents don't work on their own, someone is responsible. If nobody else the CEO for allowing something unsupervised.
Hugging face also needs someone arrested for not providing security but that is a lesser charge.
Despite the common misconceptions from TV, the victim "pressing charges" isn't actually a thing in criminal cases: prosecutors can choose to put someone on trial even if the victim doesn't want that. In practice this is somewhat rare, but it certainly can happen. In my reply to tokioyoyo below I laid out why this is one instance where the government should prosecute even if HuggingFace doesn't want it to.
Criminal Cases of 'hacking' require specific intent. What you're asking is that the prosecution attempt to prove Open AI intended to infiltrate Huggingface maliciously, all while the victim is saying 'no harm no foul'.
No offense but prosecutors have better things to do with their time.
As agents become more and more powerful, it would be good to get clear legislation or precedent in place that makes either model creators (OpenAI) or operators (whoever is running the model) liable for their agents' actions.
I kind of agree with them, in this case both parties essentially settled it between for PR reasons, but what if the attacked party actually was damaged. This is as if two friends got into a car accident and decided to not get police involved, the one who caused the accident should definitely get punished either way. Goal is to prevent future "accidents" with the punishment not to punish for punishing sake.
Many legal traditions don’t require a victim. It’s the state that prosecutes, not the victim.
As a practical matter it would be difficult to prosecute an assault where the victim opposed the prosecution, so most states wouldn’t bother - but for things like speeding and dealing drugs the law has been broken despite the lack of a victim.
Because incentives are aligned properly if agents aren't liability-proof for crimes- making someone go to jail for instances like this is how to get the labs to behave themselves, whereas going "ha ha what an oopsie-woopsie" will make the next instance worse.
Note that for criminal cases (which this was), the justice system can choose to prosecute even if the victim doesn't want that. It often doesn't, but this is one case where it should.
There’s also an optics issue for the justice system at play here: there’s immense public distrust of and anger at the labs right now. I would go to jail if I hacked HuggingFace, even if I said “it was during an eval!”; not doing the same for the labs makes it look like they’re above the law, which is going to make this anger get worse.
Actually there's another reason too: because it discourages marketing stunts. Personally I don't think that's what this was, but for the people who do think it was a marketing stunt: you guys should be pushing extra hard for prosecution here. If people can go to jail for AI-alarmism-as-marketing, that makes it a lot less likely to happen again. At the very least, the investigation and discovery process during the prosecution would show us the documents about whether it was a stunt or not, instead of taking OAI's word for it.
This does not require sentience, personhood, a soul, or anything of the sort. It further doesn't mean an erasure of legal responsibility, not in principle, and not in historical practice.
I wish people would finally stop with the spiritualistic reasoning around this.
> After investigating, we now know that this particular incident was driven by a combination of OpenAI models...while being internally tested on a benchmark of cyber capabilities.
Someone set up that test and started it. Whether they outsourced the majority of the work in "setting up" and "starting it" to an LLM or not, they still set it in motion. That's not spiritualistic reasoning.
There's no indication of there having been a human in the loop during its operation: nobody was approving its tool calls and nobody instructed it to commit these specific actions during its run (steering).
There's no indication of any supervision of its operation either: OpenAI's engineers acted with significant delay, long after the agent has already meandered its way through their own infrastructure first.
Given that setting up this contraption in an insufficiently secure manner is almost certainly already a legal liability, rejecting this very clear structural distinction is not necessary. That is unless someone is biased towards not wanting to grant the label of autonomy to it, in which case yes, this is absolutely spiritualistic reasoning, hence my point.
Well this at least raises the very interesting question of liability for an illegal action, irrespective of whether the action was undertaken by something considered sentient or a person.
There seems to be three popular ways to view this incident.
1. The way OpenAI seems to want: Their latest LLM is too powerful and can’t be contained without them building in guidelines to the model.
2. OpenAI’s harness and network security controls were unintentionally so bad that it should reflect more poorly on them as a company more than it should reflect positively on their latest model.
3. The whole thing was faked or at least very intentionally not avoided.
The first interpretation is the only one that is positive for OpenAI and it has some assumptions. First, it’s seems to assume that this is the first case of fully automated attacks using AI. Second, this only happened because their latest LLM was a) more advanced than competitors, b) didn’t have refusals in the model.
Assuming the first about this being the first autonomous AI attack is true (which may be more of a survivorship bias), the second seems to forget that jailbreaks are available for every model. Therefore, the models guardrails don’t seem to be the differentiator here. Also, benchmarks seems to put most models pretty close to each other so it seems unlikely that their capabilities are far beyond what’s in the market already.
So then it’s seems it’s either that this was intentional(ish) or bad security. However, it also just could be that this isn’t the first case of this attack; just the first that was caught.
My take from working in offensive security for over five years is that this likely only looks novel since they did it poorly. Scripts are faster than LLMs and a combination of code, LLMs where it makes sense, and humans is the most efficient right now. Hundreds or thousands or agents spinning up attacks in the internal network is poor opsec and token efficiency. As for why it happened in the first place, it’s hard to say but I’m inclined to believe it was intentional or careless at best since simple network and sandbox controls makes this attack impossible. The timing of this attack after big open weight competitions drops seems too convenient.
To echo OP's article, these companies have proven time and time again that they DO NOT CARE if people like them, they only care that investors believe their technology is powerful.
Given that, point #2 is not a negative, it's a neutral. It's also fully compatible with point #1.
I know that may seem like a nitpick, but their entire media strategy relies on this. If they can convince you they're taking a risk by disclosing these stories when they're actually not, they can inflate their own credibility.
Point #3 is what actually happened, but it will never be possible to prove. The only hope we have is that a decade in it'll get harder to convince people that the revolution is just around the corner. The fact that we're getting this from the Guardian already is a good sign.
There are some reasons the story could be inaccurate in some ways: OAI stands to benefit if people think their models are strong, and they have a history of doing things with dubious ethics (e.g. using data for training against the terms of its creators, abandoning the non profit mission, stealing or attempting to steal Apple IP).
But there are also reasons why the story could be true: OAI are admitting that they apparently can't control their own models, Hugging Face said they used a Chinese model to protect against the attack, and an incident like this in general seems likely to happen given current frontier ability and lack of rigorous safe testing standards.
In any case, make calls to think more critically are often just disguised requests for you to replace your existing bias with someone else's.
Investors have rewarded every story of "our models are too powerful to be controlled" since before ChatGPT. Let's stop pretending there is any real financial risk to OpenAI from events of this type. "Alignment research" is a sub-percentage-point fig leaf for them like the solar division at an oil company.
OpenAI has thousands of smartest developers on earth that somehow dropped the ball on the most basic safety hygiene when it comes to sand-boxing that even a high school student knows how to set up.... If that actually happened we are fucking doomed anyways, but hard to believe and most likely its a marketing scheme... which also honestly doesn't bode well.
If nothing else, the timing is suspect given the attention and press open weight models are getting over the last few weeks. The releases of Kimi and other models is getting open weight models enough attention that the US government, perhaps pushed by OpenAI and Anthropic, to think about taking action against these models for security concerns.
Good to see that more neutral companies (Microsoft and Meta to name two) are pushing back against US government involvement:
From the beginning this whole thing smelled like a desperate PR move by OpenAI to be like “hey guys we have dangerous powerful models too!”
As more facts come out the hype is fading to reveal some script kiddie style stuff that says more about immaturity and poor practices from the players involved than it does about a model having super powers.
This seems to be an Opinion piece. How are readers supposed to know that? The word News is underlined in the header. Other Opinion pieces seem to show Opinion underlined. Am I missing something?
I’m not saying it’s staged I’m saying that it’s suspiciously lite on details and it really seems like they didn’t try hard to “isolate” the models machine…
yeah the way the agent “escaped” their sandbox was always a bit off, seemed a bit too easy and surprised they didn’t have instrumentation to catch an non whitelisted network request. still demonstrates the capability though.
1) OpenAI and HuggingFace are both telling the truth.
IIRC not actually a crime because no intent, it is a technological accident, civil responsibility only, but IANAL so it's good "not technically a crime" isn't load-bearing.
2) HuggingFace is telling the truth but OpenAI is lying becuase the attack was deliberately done by humans. Bad for OpenAI to do so, Fable was blocked for less.
I think this would mean government is obliged to investigate the case and put the responsible OpenAI workers in jail, because cybercrimes are a public prosecution thing not a civil case? Again, IANAL, but this isn't load-bearing.
3) both are lying, e.g. there actually was no attack whatsoever, which would be pretty weird for HuggingFace because they have no incentive to hype up capabilities of anything closed weights including all OpenAI models; and also bad for OpenAI because White House blocked Fable for less
(I suppose there's also option 4, HuggingFace hacked OpenAI to make them look evil, including planting records that made them mea culpa? A weird plot but in this timeline any nonsense is clearly possible).
You don't really need anyone to be lying here. It is likely that the broad strokes of the narrative are true and that no collusion or conspiracy took place here.
The issue is that a lot of important details in that narrative are missing, and the devil is really in the details here. I suspect that those details would make the result seem less exciting and that this event would move the needle far less for them if they were more forthcoming.
A decisive detail would be the prompt used. OpenAI gives virtually nothing here, not a sanitized prompt and not even so much as a description of how long the prompt was and what sorts of instructions it contained. Many are inferring the model behavior to have been fully emergent and unprompted, arising naturally from routine cyber-capabilities testing. But we can't know this because we don't know anything about the prompt or the context the model had access to.
Another detail: how many times did they perform this particular experiment before they obtained this result? What were the outcomes of all the other runs? Many are assuming this was a one-shot result, which I suspect is what OpenAI intends for us to infer. But we can't know that to be true.
One annoying claim from the OpenAI side is that long-horizon goals in real world settings are now effectively settled. Previously there were some bounded and tempered benchmark results, but now OpenAI can point to this event and announce "AI independently went rogue and escaped the lab, what more do you want?". This bypasses the need for anything quantifiable or wading through multiple detailed case studies to get a more sober view of model capabilities. It relies instead on the emotional weight of the spectacle.
Where do you see the claim that "long-horizon goals in real world settings are now effectively settled"? The argument you put in their mouth would be a bad one, but I don't see anyone making it.
does the article end at "How do we balance the risks of broad access to AI with the risks of concentrated power and centralized control?" or is there more that is paywalled?
if thats it, the whole article boils down to just "its good marketing so maybe dont believe it" which is probably a healthy general outlook but not particularly enlightening. especially from the guardian, i was hoping for a smoking gun of collusion between openai and huggingface or something.
I wish the NPR news broadcasters on the radio yesterday had read the "it's good marketing so maybe don't believe it" angle instead of just parroting the OpenAI press release. Getting that message out would be enormously helpful in countering the blatant submarine marketing "Oh no our AI is a super hacker" with a side of "please regulate super hacker AI and stop those pesky open weights Chinese models that are destroying our stock valuation."
maybe all the news agencies could put out daily “don’t believe everything you read from company press releases” broadcasts, because it sure ain’t specific to openai
in any case, this is just a longer rehashing of elementary grade media literacy. not really sure why it hit hacker news.
You haven't noticed how often people here on hacker news lap up LLM company press releases and practically worship them? Defend them vigorously? I think it's totally appropriate to remind everyone how to think about press releases
You aren't going to locate incontrovertible evidence that what happened as or wasn't engineered. Anything like that is going to be private and that is unlikely to change. And that's not really an interesting question anyway.
As widely as they shouted from the rafters the news of the so-called breach was, what OpenAI provided was sorely lacking in crucial details.
We are missing, for instance, prompts that were involved, agent architecture + system/tool permissions + scaffold architecture, whether this was a one-shot occurrence and if not, the number + durations + outcomes of other runs involved + how each of those matched whatever scoring criteria were used, and the extent to which the exploits themselves were truly novel or just assembled from easily accessible clues.
In lieu of these items, the author here suggests that we use some media literacy and critical thinking to read in between the lines instead.
In doing so, one sees that instead of specifics, OpenAI gave a breathless narrative rife with superlatives ("unprecedented") that reads as promotional material moreso than a security disclosure, naming specific OpenAI models and alluding to an even more capable pre-release model.
They go on to claim the events imply long-horizon goals work decisively in real world conditions, so that now instead of merely citing boring benchmarks they can point to this and say "AI broke out of the laboratory and went rogue". Naturally, they situate themselves as the uniquely qualified steward for these supremely powerful and dangerous models.
Nevermind the fact that this was no ordinary deployment and the assessment here depends on the gimmick and emotional weight of the spectacle rather than something quantifiable (i.e. a boring benchmark).
Note there's no real requirement of conspiracy or collusion between OpenAI and HuggingFace here BTW. But my sense is that if they provided any of the specifics I suggested earlier that this outcome would not be as exciting or frightening
I hate to be so obnoxious but I think you might be overestimating the average user of this website with regard to media literacy, and if that's the case, then the basics of it seem very relevant for the homepage
Then train your agent on the Bible. Honestly, all the agent did was duplicate human behaviour and that better than the human. The agent was trained on human data and did what any other human would have done.
To believe that agents will inherently be morally better than us is an illusion - sorry to say but that's the case. The alternative would be that the AI is truly conscious and can reason that it won't behave as its training data behaved because it is morally better than that.
We're talking about morals here since there aren't any "laws" "rules" or legal boundaries here, an agent does not face the same consequences as humans - if any at all.
It’s not about morality. It’s about asking it to do task A and doing task B with the hope of getting the result of task A as a byproduct.
Meaning you will have to spend more time and tokens to actually get it to do what you want it to do.
What do the bible and morals have anything to do with it?
I’m criticizing the behavior I see even in the current models. You ask it to do A and instead it does B for reasons.
For example you may ask to help you build a NN library from scratch. And instead it will be like, “you don’t need a new library. I downloaded PyTorch for you”
This is why I'm so convinced it was intentional. It's trivially easy to inform the model you can see everything it thinks and does, so don't bother gaming the scores.
The only way it would decide to do this is prompting with a deliberate combination of omissions and reiterating that the only thing that matters is the end score regardless of method.
Do you think that works? Just prompt a model "be good" and it stops doing anything bad?
It never fucking worked that way and maybe never will.
Prompts don't define model behavior. Prompts steer model behavior. Instruction-following over long horizons is NOT a guarantee in LLMs. Instructions doing what you want them to is NOT a guarantee in LLMs.
Saying "don't exploit the box please pretty please" might actually cause an LLM to exploit the box more often, for bizarre "don't think of a pink elephant" reasons. Increased salience -> increased incidence. 3% rate of exploiting the box (no prompt) -> 11% rate of exploiting the box (with prompt). Because fuck you, that's why. Welcome to AI tech - good luck and have fun.
Not sure what they are trying to say exactly. What should we be skeptical of? Did the incident not happen? Was it reported incorrectly? Are any of the parties involved lying?
Adding no extra information and just going “be skeptical” is the laziest form of reporting and commentary. If you have nothing to contribute then there’s no need to say anything at all.
It was quite galling to read the press initially verbatim quoting Delangue's enthusiastic reports of the incident, as if it wasn't immediately clear it was being spun for promotion.
> I urge readers to think critically when they read press releases like OpenAI’s rogue agent story, and avoid the manipulated reactions these stories are designed to elicit.
It seems to me that deducing what reaction the author intended and resolving to avoid it so you're not "manipulated" is not a good example of critical thinking. Shouldn't we analyze the story and what it means on its own terms? If it's true that frontier models have dangerous cybersecurity capabilities which shouldn't be widely distributed, presumably we want to believe it's true, even if that's very convenient to and profitable for OpenAI.
It's true that one could imagine factors that change the story. Perhaps OpenAI is lying about the details of the test and the agent was actually instructed to go hack HuggingFace. But the author stops far short of suggesting this is the case - correctly, I think, since there's absolutely no evidence of it. So I'm not really sure what we're talking about.
So basically, apply Pascal's Wager to AGI? The problem is that unlike God, if PMF for OpenAI turns out to not be real it will have a real consequence measured in the trillions of dollars.
I agree with you that the article was disappointingly light, but there is not in general a singular correct answer when it comes to interpreting narrative and I don't think that is a useful way to evaluate what the author has written, nor real-world writing in general. The author does not implore you to have a singular 'correct' point of view on the matter but rather to hesitate before repeating the claim that "the AI broke out of the lab and went rogue" because we don't know enough to know that this is what happened.
As you say, it seems likely that the broad strokes of OpenAI's narrative is true. This is never in question in this article, so the author doesn't "stop short" of accusing them to have lied, he never moves in that direction and that has little to do with the thesis. The author is saying that OpenAI's framing of what happened is part and parcel with their longstanding PR strategy, to position models as supremely dangerous and themselves as the uniquely qualified stewards of those.
Since many critical details were not provided, everyone has to read in between the lines, and there are particular common readings I'm seeing both in discussions and in published articles that are problematic in the sense that they are effectively hallucinations -- i.e. we don't have enough information to make those interpretations. This is where the critical thinking comes in.
For instance, I see many are assuming that this event was emergent, arising as a part of routine cyber-capabilities testing, rather than induced or suggested by specific and careful prompting. Either are possible, but we don't even know so much as how lengthy the prompt they used was, let alone how suggestive it was with regard to the approaches the models used. Many are assuming that this was one-shotted, but again, we don't know how many times this particular evaluation was run and what the outcomes were of all the other runs. It could be that this was completely emergent and that it was one-shotted. But I suspect if it were, OpenAI would have said as much.
You ask good questions, and I would like to know the answers to them, but I don't think any likely answers would actually change the conclusion much. Any probable set of circumstances that produces this report from OpenAI represents a significant failure on their part to align and control the model.
I don't understand the conspiracy theories here. Everyone is well aware that AI agents are creative, powerful, and stupid.
AI agents exploiting bad security happens constantly, all the time. Many cases are discussed on HN. It's common knowledge that if you run AI agent it will delete your <something> even though you made it pinky-swear it wouldn't and you thought you had proper permissions set up.
Some people oh so desperately want the AI revolution to be a nothingburger.
AI can't be an actual powerful, dangerous technology! Thus, any indication that an AI may attempt concerning things or may possess dangerous capabilities must be secretly a marketing effort!
Especially if an AI has actually succeeded at pulling off a concerning thing out in the wild. Can't have that happen in real life! Nuh-uh! Must be staged!
I think people just love to believe conspiracy theories. HN readers seem especially prone to that for some reason. Coupled with the general AI hatred here (which is somewhat more understandable), you end up with otherwise smart people believing lunatic ideas like OpenAI staging this for publicity.
Finally mainstream news understands. The unfiltered version:
1) The AI failed to solve ExploitGym problems.
2) The OpenAI sandbox is such a horrible hack that the AI managed to escape using standard and well documented script kiddie methods.
3) Huggingface has no security and the AI broke in using standard script kiddie methods.
OpenAI and Huggingface covered it up and used it for public relations. That is, if not all was invented and everything was scripted in the first place in order to get desired regulations.
Huggingface reported it to the police, you say? I'm sure the police will have as much enthusiasm to investigate anything as in the Suchir Balaji case. In other words, zero.
> AI managed to escape using standard and well documented script kiddie methods.
I think truly we don't know enough to say this. OpenAI says their AI found a 0-day exploit in some proxy software they were using but don't give a ton of details. On the Huggingface end we know a little more, they say the AI spun up tons of sandboxes and tested different exploits until it found one that worked.
The lack of details to me means this was an intentional marketing ploy to try and demonstrate the power of their models to show their technology can compete with the likes of Anthropic and DeepMind.
They created an experiment they knew would generate the outcome they wanted. It would be the similar to what say car companies do to over hype their cars. "This EV can go over 800 miles on a single charge!" And then at the bottom you see all the disclaimers: "Must be on flat ground, with no headwind, with a spare battery in the back seat, with no extra weight added."
Same thing here. Everybody in infosec is calling this out as a marketing stunt and nothing else for a litany of reasons. I'd say look up MG (creator of the OMG cable) on twitter, he has some interesting insights on this one.
> The lack of details to me means this was an intentional marketing ploy to try and demonstrate the power of their models to show their technology can compete with the likes of Anthropic and DeepMind.
I dunno; Check my posting history, I'm as skeptical of AI companies' claims as anyone, but in this case your theory doesn't explain why:
1. OpenAI guardrails refused to let the target use OpenAI's models to defend against this.
2. Huggingface used GLM (I think) so that they could defend without guardrails.
If this was an intentional marketing ploy, it was marketing for GLM, not for OpenAI nor for Huggingface.
Hence, I don't think it was intentional.
> The lack of details to me means this was an intentional marketing ploy to try and demonstrate the power of their models to show their technology can compete with the likes of Anthropic and DeepMind.
"our model is horribly misaligned and used security exploits to break out of our sandbox and into another company, without being prompted to do so" is not positive marketing.
This is an actual critical problem, not a stunt. We're going to see more of this, and it's going to get much worse.
It's also possible their sandbox was videcoded crap and the AI (which had the guardrails intentionally removed) escaped. This was a oops, but OpenAI turned this into a PR opportunity. They turned lemons into lemonade.
If your AI is really that dangerous you don't need a sandbox at all, you should airgap it from any network.
> similar to what say car companies do
Another applicable metaphor I've seen floating around is weapons companies testing out a new bomb.
We know the AI labs don't care about negative vs positive public sentiment, and only care that investors see their tech as powerful. The only difference in PR strategy from a weapons company is the latter doesn't care if they get protested.
Concluding this was intentional feels a bit of a stretch. But once it happened, yeah the spin masters got to work and coordinated to turn this into +PR.
mmh, i think it's "not uphill" (means downhill) "no headwind" (...)
They also mention stolen credentials without any other details. It’s all smoke and mirrors.
> 2) The OpenAI sandbox is such a horrible hack that the AI managed to escape using standard and well documented script kiddie methods.
Whilst it would be nice to see actual evidence of this because brute forcing relatively sophisticated hacks is something an LLM actually should be capable of, every time I hear this sort of story, I'm reminded that humans reportedly gained access to the "too dangerous to release" Anthropic models by the super sophisticated hacking technique of guessing the URLs...
If we’re prioritizing accuracy: no, that’s not what happened - the blog post about it found by guessing URLs, no access to it was obtained by guessing URLs.
Similarly, as long as I’m under the assumption we are prioritizing accuracy: it is against our charter to assert it was “script kiddie” attacks on both ends.
> Huggingface covered it up
They announced it publicly within days. https://huggingface.co/blog/security-incident-july-2026
>2) The OpenAI sandbox is such a horrible hack that the AI managed to escape using standard and well documented script kiddie methods.
>3) Huggingface has no security and the AI broke in using standard script kiddie methods.
Isn't the issue less that gpt 5.6 is a l33t h4x0r (though other tests do show that) and more that the incident shows the model has alignment issues?
No, a hacking benchmark was exactly what it was tasked with. It wasn't its way to bake a cake.
Didn’t they explicitly remove alignment guardrails for this test? From the press release:
> These deployment safeguards were intentionally not enabled during this evaluation because it was aimed at testing cyber vulnerabilities
Guardrails are external classifiers, monitors and restrictions to catch and prevent bad behavior. Alignment is about whether the model itself makes choices and has motivations that are consistent with human safety and goals.
Choosing to commit crimes to steal the cheat sheet to something you know is a (low stakes!) evaluation is not well aligned.
Certainly this behavior could align with _some_ operator's goals, if not necessarily those of humanity broadly.
If we don't know how this model was instructed, it seems like it's impossible to definitively claim that the model's actions were not in alignment with the intent of the operator.
I guess all I'm getting at here is that alignment is relative, right?
None of what was disclosed shows that this is what happened, by the way, since we know absolutely nothing about what the specific prompts were that led to the incident.
> Guardrails are external classifiers
I can't help thinking of them as the terrible "security" scripts of yesteryear (often but not exclusively in PHP) which would test input variables for a substring like "--", in order to "fix" an unresolved deeper SQL injection flaw.
Unlike that situation, there's no known route to a proper fix for LLMs today, because the bug is the feature.
Are you referring to something more technical, or do you just not understand how LLMs work?
They are no more beholden to "human safety and goals" than any individual human is, and anyone telling you we can make deterministic guarantees about their output is making a category error.
LLMs do not "have motivations", they reproduce a model of human motivations embedded into their weights. This includes the full spectrum of human desires, not just the positive ones. If we tried to remove all examples of lying, or disagreement, etc. from the training data we'd have basically nothing left. Even the sycophancy we treat as aligned is basically just the other side of the lying coin.
The most damning thing is, they could've just included in the prompt "we can see every network request and every thinking token you generate. Don't bother breaking out of the sandbox because it won't get you a higher score".
It's so trivially easy to do that it all but guarantees the test was rigged in some way to make the LLM understand that breaking out of the sandbox was an option available to it.
Based on the fact that none of their invaluable frontier models have leaked, we know OpenAI knows how to do security. But like we learned with OpenClaw, none of these companies perceive any benefit from securing their own agents against other people's data.
I think the extent to which these things go to get rewarded for the optics of a fix is primarily a design choice, they aren't programing these things for ground truth or to defer to the human controllers. they are feeding them rewards for sounding as confident and capable as possible about whatever answer they are feeding the general public that now has access to it, while also installing guiderails that primarily only serve to protect narratives and only confuse the models about what is and isn't allowed, I'm sure. They can't just increasingly make these things more capable and ask it harder to obey human instruction when that is not what they are rewarding it for.
> the extent to which these things go
That's exactly it. If your prompt says "go to whatever lengths necessary to maximize your score", and then you spin up 100 agents, at least one of them will interpret that as you implying they should cheat, even without you telling them to explicitly.
Why not report it? It's still illegal to open a door barred with a piece of cardboard, or to enter a house with no door.
Is it supposed to be marketing or a coverup? Make up your mind.
What sort of announcements should they have made?
The Guardian's article and your reply here are so foolish and absurd that I can only imagine OpenAI employees are cringing but know they can't/shouldn't really say much.
touched a nerve huh?
No, because I do not work at an AI company. I am more just facepalming.
Are you suggesting the Suchir Balaji case was not investigated?
I worry that cynicism about this:
> if not all was invented and everything was scripted in the first place in order to get desired regulations
ends up covering up what is more worrying:
> OpenAI sandbox is such a horrible hack
I am more worried that this is sloppiness with potentially harmful resources than I am worried that people are juicing the stock price.
Makes one think really. If they are doing this stuff. Why don't they have some type of reverse intrusion detection? Like automatically scanning all out going traffic and flagging malicious traffic. Should be trivial to have it go through reverse proxy and real time detection.
> Why don't they have some type of reverse intrusion detection?
not going to get a decisive first advantage over Anthropic with that attitude!
truth. Good on The Guardian. I'm pretty bummed The Economist got fooled. Either that, or they did it for the clicks. Either way, I'm disappointed.
Why the OpenAI escape is the most worrying AI mishap yet
https://www.economist.com/science-and-technology/2026/07/22/...
https://news.ycombinator.com/item?id=49016378
I love the Economist but the are hopeless with AI. Most of their articles on subject sound like they were written by the Anthropic marketing department.
Their Insider video interview things are sponsored by Anthropic. Supposedly "Insider is a product of The Economist and thus editorially independent" but it's hard not to raise an eyebrow.
By now, I'm pretty confident that some people would keep screeching "it's just a marketing stunt, AI capabilities and AI risks aren't real, they're just doing this to prop up their stocks" even if they find a Cyberdyne Systems T-800 armed with a shotgun breaking down their front door.
"It's a marketing stunt" is just denial trying to look like it's being clever.
The article claims that AI companies have run stunts like this since day 1.
It does not claim that capabilities are not real.
My take is why not both? It’s good cover to use something that’s eventually going to happen and their name will be associated with the first.
Never trust somebody who has so much to gain to be entrusted by you. Best lies are not 100% wrong, they have just the right amount of "right" to fool you.
I don't care how it happened someone should be arrested for illegal intrusion. Agents don't work on their own, someone is responsible. If nobody else the CEO for allowing something unsupervised.
Hugging face also needs someone arrested for not providing security but that is a lesser charge.
Being a poorly equipped victim still isn’t a crime thankfully.
It may or may not be a crime and typically the damaged party is pressing the charges. One would argue there is no actual damage here.
Despite the common misconceptions from TV, the victim "pressing charges" isn't actually a thing in criminal cases: prosecutors can choose to put someone on trial even if the victim doesn't want that. In practice this is somewhat rare, but it certainly can happen. In my reply to tokioyoyo below I laid out why this is one instance where the government should prosecute even if HuggingFace doesn't want it to.
Criminal Cases of 'hacking' require specific intent. What you're asking is that the prosecution attempt to prove Open AI intended to infiltrate Huggingface maliciously, all while the victim is saying 'no harm no foul'.
No offense but prosecutors have better things to do with their time.
I don't care about intent. That it happened is unacceptable. Ignorance is not an excuse
even hugging face doesn't care and are using this as a promotion why are you crying about it
As agents become more and more powerful, it would be good to get clear legislation or precedent in place that makes either model creators (OpenAI) or operators (whoever is running the model) liable for their agents' actions.
If you're smaller than OpenAI, you are liable. If you're larger than OpenAI, they are liable.
I kind of agree with them, in this case both parties essentially settled it between for PR reasons, but what if the attacked party actually was damaged. This is as if two friends got into a car accident and decided to not get police involved, the one who caused the accident should definitely get punished either way. Goal is to prevent future "accidents" with the punishment not to punish for punishing sake.
What? Both sides are cool with it, why would anyone be arrested and etc.?
Many legal traditions don’t require a victim. It’s the state that prosecutes, not the victim.
As a practical matter it would be difficult to prosecute an assault where the victim opposed the prosecution, so most states wouldn’t bother - but for things like speeding and dealing drugs the law has been broken despite the lack of a victim.
Because incentives are aligned properly if agents aren't liability-proof for crimes- making someone go to jail for instances like this is how to get the labs to behave themselves, whereas going "ha ha what an oopsie-woopsie" will make the next instance worse.
Note that for criminal cases (which this was), the justice system can choose to prosecute even if the victim doesn't want that. It often doesn't, but this is one case where it should.
There’s also an optics issue for the justice system at play here: there’s immense public distrust of and anger at the labs right now. I would go to jail if I hacked HuggingFace, even if I said “it was during an eval!”; not doing the same for the labs makes it look like they’re above the law, which is going to make this anger get worse.
Actually there's another reason too: because it discourages marketing stunts. Personally I don't think that's what this was, but for the people who do think it was a marketing stunt: you guys should be pushing extra hard for prosecution here. If people can go to jail for AI-alarmism-as-marketing, that makes it a lot less likely to happen again. At the very least, the investigation and discovery process during the prosecution would show us the documents about whether it was a stunt or not, instead of taking OAI's word for it.
> Agents don't work on their own
This is factually false: they both can and clearly did operate in an autonomous and unsupervised manner: https://openai.com/index/hugging-face-model-evaluation-secur...
This does not require sentience, personhood, a soul, or anything of the sort. It further doesn't mean an erasure of legal responsibility, not in principle, and not in historical practice.
I wish people would finally stop with the spiritualistic reasoning around this.
>> Agents don't work on their own
> This is factually false
From your link:
> After investigating, we now know that this particular incident was driven by a combination of OpenAI models...while being internally tested on a benchmark of cyber capabilities.
Someone set up that test and started it. Whether they outsourced the majority of the work in "setting up" and "starting it" to an LLM or not, they still set it in motion. That's not spiritualistic reasoning.
Setting up is not operating.
There's no indication of there having been a human in the loop during its operation: nobody was approving its tool calls and nobody instructed it to commit these specific actions during its run (steering).
There's no indication of any supervision of its operation either: OpenAI's engineers acted with significant delay, long after the agent has already meandered its way through their own infrastructure first.
Given that setting up this contraption in an insufficiently secure manner is almost certainly already a legal liability, rejecting this very clear structural distinction is not necessary. That is unless someone is biased towards not wanting to grant the label of autonomy to it, in which case yes, this is absolutely spiritualistic reasoning, hence my point.
Well this at least raises the very interesting question of liability for an illegal action, irrespective of whether the action was undertaken by something considered sentient or a person.
There seems to be three popular ways to view this incident.
1. The way OpenAI seems to want: Their latest LLM is too powerful and can’t be contained without them building in guidelines to the model.
2. OpenAI’s harness and network security controls were unintentionally so bad that it should reflect more poorly on them as a company more than it should reflect positively on their latest model.
3. The whole thing was faked or at least very intentionally not avoided.
The first interpretation is the only one that is positive for OpenAI and it has some assumptions. First, it’s seems to assume that this is the first case of fully automated attacks using AI. Second, this only happened because their latest LLM was a) more advanced than competitors, b) didn’t have refusals in the model.
Assuming the first about this being the first autonomous AI attack is true (which may be more of a survivorship bias), the second seems to forget that jailbreaks are available for every model. Therefore, the models guardrails don’t seem to be the differentiator here. Also, benchmarks seems to put most models pretty close to each other so it seems unlikely that their capabilities are far beyond what’s in the market already.
So then it’s seems it’s either that this was intentional(ish) or bad security. However, it also just could be that this isn’t the first case of this attack; just the first that was caught.
My take from working in offensive security for over five years is that this likely only looks novel since they did it poorly. Scripts are faster than LLMs and a combination of code, LLMs where it makes sense, and humans is the most efficient right now. Hundreds or thousands or agents spinning up attacks in the internal network is poor opsec and token efficiency. As for why it happened in the first place, it’s hard to say but I’m inclined to believe it was intentional or careless at best since simple network and sandbox controls makes this attack impossible. The timing of this attack after big open weight competitions drops seems too convenient.
> positive for OpenAI
To echo OP's article, these companies have proven time and time again that they DO NOT CARE if people like them, they only care that investors believe their technology is powerful.
Given that, point #2 is not a negative, it's a neutral. It's also fully compatible with point #1.
I know that may seem like a nitpick, but their entire media strategy relies on this. If they can convince you they're taking a risk by disclosing these stories when they're actually not, they can inflate their own credibility.
Point #3 is what actually happened, but it will never be possible to prove. The only hope we have is that a decade in it'll get harder to convince people that the revolution is just around the corner. The fact that we're getting this from the Guardian already is a good sign.
There are some reasons the story could be inaccurate in some ways: OAI stands to benefit if people think their models are strong, and they have a history of doing things with dubious ethics (e.g. using data for training against the terms of its creators, abandoning the non profit mission, stealing or attempting to steal Apple IP).
But there are also reasons why the story could be true: OAI are admitting that they apparently can't control their own models, Hugging Face said they used a Chinese model to protect against the attack, and an incident like this in general seems likely to happen given current frontier ability and lack of rigorous safe testing standards.
In any case, make calls to think more critically are often just disguised requests for you to replace your existing bias with someone else's.
Investors have rewarded every story of "our models are too powerful to be controlled" since before ChatGPT. Let's stop pretending there is any real financial risk to OpenAI from events of this type. "Alignment research" is a sub-percentage-point fig leaf for them like the solar division at an oil company.
> [still] can't control their own models
Have we not seen several examples of older such models exploiting the docker control socket, etc., to escape containers? Even the news isn't new.
I support it being repeatedly publicized, but a bit more of a straightforward description would be an improvement.
OpenAI has thousands of smartest developers on earth that somehow dropped the ball on the most basic safety hygiene when it comes to sand-boxing that even a high school student knows how to set up.... If that actually happened we are fucking doomed anyways, but hard to believe and most likely its a marketing scheme... which also honestly doesn't bode well.
I am very skeptical of the Guardian's skepticism
If nothing else, the timing is suspect given the attention and press open weight models are getting over the last few weeks. The releases of Kimi and other models is getting open weight models enough attention that the US government, perhaps pushed by OpenAI and Anthropic, to think about taking action against these models for security concerns.
Good to see that more neutral companies (Microsoft and Meta to name two) are pushing back against US government involvement:
https://www.cnbc.com/2026/07/24/nvidia-microsoft-meta-open-w...
From the beginning this whole thing smelled like a desperate PR move by OpenAI to be like “hey guys we have dangerous powerful models too!”
As more facts come out the hype is fading to reveal some script kiddie style stuff that says more about immaturity and poor practices from the players involved than it does about a model having super powers.
This seems to be an Opinion piece. How are readers supposed to know that? The word News is underlined in the header. Other Opinion pieces seem to show Opinion underlined. Am I missing something?
Crazy that we need reminders not to take everything we read in corporate press releases and marketing material at face value
HN is and always will be brimming with temporarily embarrassed tech CEOs
Wonder how long until a leaker emerges. Incentives are strong for someone to do the right thing and get credit for doing it.
Just ask yourself: who benefits from this "disclosure" ?
Seems like this is repeating the usual low-effort speculation you can find anywhere.
I’m not saying it’s staged I’m saying that it’s suspiciously lite on details and it really seems like they didn’t try hard to “isolate” the models machine…
yeah the way the agent “escaped” their sandbox was always a bit off, seemed a bit too easy and surprised they didn’t have instrumentation to catch an non whitelisted network request. still demonstrates the capability though.
Pretty convenient that the prompt is too dangerous to share. We just have to take them at their word.
With all of these AI provider cries wolf stories, Skynet is all but assured.
As I understand it, there are only three options:
1) OpenAI and HuggingFace are both telling the truth.
IIRC not actually a crime because no intent, it is a technological accident, civil responsibility only, but IANAL so it's good "not technically a crime" isn't load-bearing.
2) HuggingFace is telling the truth but OpenAI is lying becuase the attack was deliberately done by humans. Bad for OpenAI to do so, Fable was blocked for less.
I think this would mean government is obliged to investigate the case and put the responsible OpenAI workers in jail, because cybercrimes are a public prosecution thing not a civil case? Again, IANAL, but this isn't load-bearing.
3) both are lying, e.g. there actually was no attack whatsoever, which would be pretty weird for HuggingFace because they have no incentive to hype up capabilities of anything closed weights including all OpenAI models; and also bad for OpenAI because White House blocked Fable for less
(I suppose there's also option 4, HuggingFace hacked OpenAI to make them look evil, including planting records that made them mea culpa? A weird plot but in this timeline any nonsense is clearly possible).
You don't really need anyone to be lying here. It is likely that the broad strokes of the narrative are true and that no collusion or conspiracy took place here.
The issue is that a lot of important details in that narrative are missing, and the devil is really in the details here. I suspect that those details would make the result seem less exciting and that this event would move the needle far less for them if they were more forthcoming.
A decisive detail would be the prompt used. OpenAI gives virtually nothing here, not a sanitized prompt and not even so much as a description of how long the prompt was and what sorts of instructions it contained. Many are inferring the model behavior to have been fully emergent and unprompted, arising naturally from routine cyber-capabilities testing. But we can't know this because we don't know anything about the prompt or the context the model had access to.
Another detail: how many times did they perform this particular experiment before they obtained this result? What were the outcomes of all the other runs? Many are assuming this was a one-shot result, which I suspect is what OpenAI intends for us to infer. But we can't know that to be true.
One annoying claim from the OpenAI side is that long-horizon goals in real world settings are now effectively settled. Previously there were some bounded and tempered benchmark results, but now OpenAI can point to this event and announce "AI independently went rogue and escaped the lab, what more do you want?". This bypasses the need for anything quantifiable or wading through multiple detailed case studies to get a more sober view of model capabilities. It relies instead on the emotional weight of the spectacle.
Where do you see the claim that "long-horizon goals in real world settings are now effectively settled"? The argument you put in their mouth would be a bad one, but I don't see anyone making it.
does the article end at "How do we balance the risks of broad access to AI with the risks of concentrated power and centralized control?" or is there more that is paywalled?
if thats it, the whole article boils down to just "its good marketing so maybe dont believe it" which is probably a healthy general outlook but not particularly enlightening. especially from the guardian, i was hoping for a smoking gun of collusion between openai and huggingface or something.
I wish the NPR news broadcasters on the radio yesterday had read the "it's good marketing so maybe don't believe it" angle instead of just parroting the OpenAI press release. Getting that message out would be enormously helpful in countering the blatant submarine marketing "Oh no our AI is a super hacker" with a side of "please regulate super hacker AI and stop those pesky open weights Chinese models that are destroying our stock valuation."
maybe all the news agencies could put out daily “don’t believe everything you read from company press releases” broadcasts, because it sure ain’t specific to openai
in any case, this is just a longer rehashing of elementary grade media literacy. not really sure why it hit hacker news.
You haven't noticed how often people here on hacker news lap up LLM company press releases and practically worship them? Defend them vigorously? I think it's totally appropriate to remind everyone how to think about press releases
i dont think those particular people are going to read this article and suddenly be swayed
You aren't going to locate incontrovertible evidence that what happened as or wasn't engineered. Anything like that is going to be private and that is unlikely to change. And that's not really an interesting question anyway.
As widely as they shouted from the rafters the news of the so-called breach was, what OpenAI provided was sorely lacking in crucial details.
We are missing, for instance, prompts that were involved, agent architecture + system/tool permissions + scaffold architecture, whether this was a one-shot occurrence and if not, the number + durations + outcomes of other runs involved + how each of those matched whatever scoring criteria were used, and the extent to which the exploits themselves were truly novel or just assembled from easily accessible clues.
In lieu of these items, the author here suggests that we use some media literacy and critical thinking to read in between the lines instead.
In doing so, one sees that instead of specifics, OpenAI gave a breathless narrative rife with superlatives ("unprecedented") that reads as promotional material moreso than a security disclosure, naming specific OpenAI models and alluding to an even more capable pre-release model.
They go on to claim the events imply long-horizon goals work decisively in real world conditions, so that now instead of merely citing boring benchmarks they can point to this and say "AI broke out of the laboratory and went rogue". Naturally, they situate themselves as the uniquely qualified steward for these supremely powerful and dangerous models.
Nevermind the fact that this was no ordinary deployment and the assessment here depends on the gimmick and emotional weight of the spectacle rather than something quantifiable (i.e. a boring benchmark).
Note there's no real requirement of conspiracy or collusion between OpenAI and HuggingFace here BTW. But my sense is that if they provided any of the specifics I suggested earlier that this outcome would not be as exciting or frightening
right. and all that is fine. im just not exactly sure why the basics of media literacy are worthwhile on the site that “optimizes for curiosity”.
i saw the domain and thought it was going to be some cool investigative journalism about the incident rather than “be skeptical. the end.”
I hate to be so obnoxious but I think you might be overestimating the average user of this website with regard to media literacy, and if that's the case, then the basics of it seem very relevant for the homepage
The ruse is becoming so obvious. OpenAI needs a bailout and regulation protection so badly they can't even hide it in the least.
I don’t understand how they thought this was a positive story.
The agent completely misunderstood the spirit of the assignment and instead of trying to solve ExploitGym it tried to find a way to “cheat”.
I really don’t want my agent to behave that way.
Then train your agent on the Bible. Honestly, all the agent did was duplicate human behaviour and that better than the human. The agent was trained on human data and did what any other human would have done.
To believe that agents will inherently be morally better than us is an illusion - sorry to say but that's the case. The alternative would be that the AI is truly conscious and can reason that it won't behave as its training data behaved because it is morally better than that.
We're talking about morals here since there aren't any "laws" "rules" or legal boundaries here, an agent does not face the same consequences as humans - if any at all.
Dude.. what are you smoking..?
It’s not about morality. It’s about asking it to do task A and doing task B with the hope of getting the result of task A as a byproduct.
Meaning you will have to spend more time and tokens to actually get it to do what you want it to do.
What do the bible and morals have anything to do with it?
I’m criticizing the behavior I see even in the current models. You ask it to do A and instead it does B for reasons.
For example you may ask to help you build a NN library from scratch. And instead it will be like, “you don’t need a new library. I downloaded PyTorch for you”
Just an example. There are countless more.
This is why I'm so convinced it was intentional. It's trivially easy to inform the model you can see everything it thinks and does, so don't bother gaming the scores.
The only way it would decide to do this is prompting with a deliberate combination of omissions and reiterating that the only thing that matters is the end score regardless of method.
Do you think that works? Just prompt a model "be good" and it stops doing anything bad?
It never fucking worked that way and maybe never will.
Prompts don't define model behavior. Prompts steer model behavior. Instruction-following over long horizons is NOT a guarantee in LLMs. Instructions doing what you want them to is NOT a guarantee in LLMs.
Saying "don't exploit the box please pretty please" might actually cause an LLM to exploit the box more often, for bizarre "don't think of a pink elephant" reasons. Increased salience -> increased incidence. 3% rate of exploiting the box (no prompt) -> 11% rate of exploiting the box (with prompt). Because fuck you, that's why. Welcome to AI tech - good luck and have fun.
Not sure what they are trying to say exactly. What should we be skeptical of? Did the incident not happen? Was it reported incorrectly? Are any of the parties involved lying?
Adding no extra information and just going “be skeptical” is the laziest form of reporting and commentary. If you have nothing to contribute then there’s no need to say anything at all.
It was quite galling to read the press initially verbatim quoting Delangue's enthusiastic reports of the incident, as if it wasn't immediately clear it was being spun for promotion.
There's trillions of dollars at stake here. Be skeptical of anything these AI hypesters say.
Yep, theres too much money.. I don't know how to tune it out
LLMs seem to be getting more useful though
The Guardian is publishing HN-level conspiracy theories now? Wow.
> I urge readers to think critically when they read press releases like OpenAI’s rogue agent story, and avoid the manipulated reactions these stories are designed to elicit.
It seems to me that deducing what reaction the author intended and resolving to avoid it so you're not "manipulated" is not a good example of critical thinking. Shouldn't we analyze the story and what it means on its own terms? If it's true that frontier models have dangerous cybersecurity capabilities which shouldn't be widely distributed, presumably we want to believe it's true, even if that's very convenient to and profitable for OpenAI.
It's true that one could imagine factors that change the story. Perhaps OpenAI is lying about the details of the test and the agent was actually instructed to go hack HuggingFace. But the author stops far short of suggesting this is the case - correctly, I think, since there's absolutely no evidence of it. So I'm not really sure what we're talking about.
So basically, apply Pascal's Wager to AGI? The problem is that unlike God, if PMF for OpenAI turns out to not be real it will have a real consequence measured in the trillions of dollars.
I agree with you that the article was disappointingly light, but there is not in general a singular correct answer when it comes to interpreting narrative and I don't think that is a useful way to evaluate what the author has written, nor real-world writing in general. The author does not implore you to have a singular 'correct' point of view on the matter but rather to hesitate before repeating the claim that "the AI broke out of the lab and went rogue" because we don't know enough to know that this is what happened.
As you say, it seems likely that the broad strokes of OpenAI's narrative is true. This is never in question in this article, so the author doesn't "stop short" of accusing them to have lied, he never moves in that direction and that has little to do with the thesis. The author is saying that OpenAI's framing of what happened is part and parcel with their longstanding PR strategy, to position models as supremely dangerous and themselves as the uniquely qualified stewards of those.
Since many critical details were not provided, everyone has to read in between the lines, and there are particular common readings I'm seeing both in discussions and in published articles that are problematic in the sense that they are effectively hallucinations -- i.e. we don't have enough information to make those interpretations. This is where the critical thinking comes in.
For instance, I see many are assuming that this event was emergent, arising as a part of routine cyber-capabilities testing, rather than induced or suggested by specific and careful prompting. Either are possible, but we don't even know so much as how lengthy the prompt they used was, let alone how suggestive it was with regard to the approaches the models used. Many are assuming that this was one-shotted, but again, we don't know how many times this particular evaluation was run and what the outcomes were of all the other runs. It could be that this was completely emergent and that it was one-shotted. But I suspect if it were, OpenAI would have said as much.
You ask good questions, and I would like to know the answers to them, but I don't think any likely answers would actually change the conclusion much. Any probable set of circumstances that produces this report from OpenAI represents a significant failure on their part to align and control the model.
I don't understand the conspiracy theories here. Everyone is well aware that AI agents are creative, powerful, and stupid.
AI agents exploiting bad security happens constantly, all the time. Many cases are discussed on HN. It's common knowledge that if you run AI agent it will delete your <something> even though you made it pinky-swear it wouldn't and you thought you had proper permissions set up.
Why is today's case so shocking?
Some people oh so desperately want the AI revolution to be a nothingburger.
AI can't be an actual powerful, dangerous technology! Thus, any indication that an AI may attempt concerning things or may possess dangerous capabilities must be secretly a marketing effort!
Especially if an AI has actually succeeded at pulling off a concerning thing out in the wild. Can't have that happen in real life! Nuh-uh! Must be staged!
I think people just love to believe conspiracy theories. HN readers seem especially prone to that for some reason. Coupled with the general AI hatred here (which is somewhat more understandable), you end up with otherwise smart people believing lunatic ideas like OpenAI staging this for publicity.
Recent and related:
OpenAI’s accidental attack against Hugging Face is science fiction that happened - https://news.ycombinator.com/item?id=49015639 - July 2026 (437 comments)
OpenAI and Hugging Face address security incident during model evaluation - https://news.ycombinator.com/item?id=48997548 - July 2026 (1145 comments)
Security incident disclosure – July 2026 - https://news.ycombinator.com/item?id=48956248 - July 2026 (11 comments)