I think the most obvious chicken little sign has been that we've had no catastrophic exploits in the past year due these findings.

Sure LLMs have helped reduce the attack surface and we patch things faster too, but I was really expected a few more Supermarket-level bangers.

There was that PHP thing, as well as *sshole, and how do we know the supermarket attackers didn't use AI?

But PHP thing was a lot of drama over something that was not super large easy to exploit hole. Just a normal "have to patch it and bye" medium kind issue.

But imagine the potential fallout if so many services didn’t get so scared that they preemptively took themselves offline.

Apply sarcasm or not based on sleep deprivation.

People say this vulnerability was "not super large easy to exploit" because of one poorly-informed skeptical blog post that hit the top of HN, but the evidence strongly suggests that post was simply wrong, and we're just going to have to live with the misinformation until it decays.

At this point the play is simply to move the goalposts, suggesting that vulnerabilities don't matter so long as you yourself couldn't easily exploit them, regardless of what well-armed attackers can do with them. That's, obviously, not how software security works.

Are there better sources for understanding the scope of impact enabled by PHP2Shell? My impression from a variety of sources is that very few incidents can be clearly attributed to that vulnerability. As opposed the many other ol’ ways PHP applications get exploited.

> the evidence strongly suggests that post was simply wrong

What sort of evidence? What "strongly suggests" this was actually easy to exploit?

I'm not interested in arguing with you about whether e.g. Dan Guido is a credible source of exploitation information.

What a smug reply. Maybe you could just link to what Dan Guido has to say about it that supports your statement and why people should find that credible.

The fact that "steal my bitcoin" got its bitcoin stolen within 2 hours?

It's quite difficult to exploit. But so are most vulnerabilities, but attackers have figured out frameworks for exploiting hard vulnerabilities of this type. The proof is in the pudding - "steal my bitcoin" had its bitcoin stolen.

I think the post scan curl post did a lot of damage to the perception of real dangers.

Instead of highlighting the exceptional quality of curl, under intense, unprecedented scrutiny, the maintainer chose to dismiss Mythos as a nothingburger. I think this was a huge mistake and a disservice to curl itself.

The goalpost here was an apocalypse, thousands instantly hacked sites and hundreds of exploits flying around within hours. I did not made that goalpost. And reality did not hit that goalpost.

> vulnerabilities don't matter so long as you yourself couldn't easily exploit them, regardless of what well-armed attackers can do with them.

This is moving goalpost. The goalpost was not "yet another vulnerability like the others".

> Just a normal "have to patch it and bye" medium kind issue.

Patching it was the issue, not the hole.

1. The patches had to be unofficially backported to the versions of php that most people were using.

2. Devs had to figure out if the unofficial backports were safe and they weren't just falling into trapped versions

3. Devs had to migrate all of their projects to unofficial versions

4. No one was ready for it to happen because there was no advance notice (someone broke the embargo) for everyone to plan maintenance windows and upgrade at the same time.

While I'm not going to claim the PHP thing wasn't a real vulnerability, it... doesn't seem to have actually been catastrophically exploited? Maybe it's just too soon and over the next year we'll see a steady stream of people announcing leaks caused by failing to update php in time? Maybe there's a bunch of servers out there running old version of wordpress and quietly mining bitcoin because of it?

It wasn't exploited because half the internet shut down until it was fixed. Someone put up a site called "steal my bitcoin" with a vulnerable version of PHP and their bitcoin private key on the same server. It was hacked within two hours (of actually enabling PHP on the server) by someone with an AI agent.

19.5hrs, not 2 hours.

https://stealmybitcoin.com/

We've had catastrophic vulnerabilities revealed every other day for months, what are you talking about?

Vulnerabilities and exploits are different things.

You mean like Notion being breached, refrigerators being turned off, that kind of thing?

Those are exploits. They are different from vulnerabilities.

Yeah, I know, and we've had a deluge of both lately. I'm not sure what constructive point you're making.

[delayed]

Assume the original message said "exploits" in place of "vulnerabilities". The point stands, independent of pedantic terminology corrections.

The correction is valid. It would be better if people used the right term. It's ok to correct. But it would be better if it were in addition to discussion of the actual point, not instead of it.

(And the phrase "catastrophic vulnerabilities" is pretty obviously talking about exploits anyway.)

> The point stands, independent of pedantic terminology corrections

the "pedantic" difference is between something being exploited and not being exploited

> Assume the original message said "exploits" in place of "vulnerabilities".

if the original message said something else, then it would mean something else.

But we've had both. Soooo?

Industrial coolers and boilers and so forth being publicly exposed isn't a new thing though, people just used to avoid targeting civilian infrastructure in cyberwarfare... That's not like a new AI thing that's just human stupidity on display...

I would hardly say hardly been literally every other day. That’s incredibly hyperbolic

Given how fast progress has been in the last year, be careful what you wish for...

For example, as others have noted, there were no results on cryptography in OpenAI's October 6th math drop. But it seems unlikely that they didn't look at this sort of thing.

That seems like a damnned if you do and damnned if you don't situation.

There were conflicting claims about the degree to which the $6 Billion Binance heist was an AI thing. But either way it seems like it would be a "look what it enabled" or a look what it failed to prevent"

If nothing happens at all it's waved away as scaremongering rather than prevention.

I don't understand, it seems like every week there is a new major breach. Multiple law firms have had unprecedented breaches, dev platforms, and companies of every kind. You name it, there's been a high profile breach.

At the point I wonder if there is any information left secure honestly. These are not from glasswing, but they are from LLMs.

There's an anecdote about curl, but I'm not sure if there's enough in here to draw conclusions from. 20% of vulnerabilities have been made public, and many of those are marked as "withdrawn" or "duplicates". However, it's unclear whether the remaining 80% have not been made public because they're not fixed or real. In other words, this is analysis of survivors (public); not much can be said about those still under embargo.

And on top of that, it's a bit negative. How many millions of dollars is one perfect-10 scoring CVSS finding, total destruction style RCE, found in curl worth? Considering how many people use it.

It's worth zero millions dollars to curl. It might be worth something to attackers, but "wide use" is not the same as "the ability to compromise everyone" because most of the time, curl isn't used to automatically fetch untrusted URLs in a way that's controlled by the attacker from an environment they care about. It's worth much, much less than a 0-day in an iPhone or PHP. You'd need a buyer with a very specific itch to scratch.

By extension, if a vuln is worth something to attackers, it's probably also worth some fraction of it to the people the attackers may go after. But if you're an interesting target, your defense can't be based on the ability to find about every single software defect ahead of the time. You need layers. And if you have that, the importance of proactively acquiring every 0-day is diminished.

Your layers need to be secure too, or you just have colanders stacked together.

A client-side bug in a command line utility that is mostly used as a debugging tool or glue in shell scripts where the URL is not attacker controlled? Probably close to nothing. Maybe if you can find some other software which lets you run curl with an attacker controlled URL, you could package it up as an exploit for that piece of software, but it's not a common config.

Curl is also a library. Just saying :-).

Fair, libcurl is pretty widely used in PHP, but it has to be in the core.

Still way less impactful than the PHP2Shell issue.

libcurl is used in a lot more than PHP. libcurl is installed on more than 20 billion devices.

... 50% of which probably never get connected to the internet, and 50% of the remainder probably only connect to one hardcoded domain name.

I'd expect the remaining 80% are delayed not because there's something statistically special about them, but from the massive workload thrown at humans in the loop.

Anthropic needs to teach their systems to propose a fix too. Detecting a coding problem and proposing a patch is what we should expect from a company with supposedly the best coding LLM.

Claude is perfectly capable of fixing these vulnerabilities; if that's what maintainers want, they can just ask for it. If the concern is cost, Anthropic offers a Claude for OSS program.

The reason Anthropic isn't proactively sending patches is, I suspect, that those patches would not be welcomed, because trust in the models isn't currently at the point where maintainers are willing to merge them without human review, which takes approximately as much effort as writing the patches themselves.

Anthropic's oss-scanner includes patches. We've received a batch and I'm working on some of it.

A good amount of the patches is straight garbage. It's not helped by the fact that a good amount (20~30%, maybe more) of the issues is also junk themselves.

I'll hopefully be able to share numbers (or data) after we're done with the batch.

70-80% not being garbage sounds like a good thing :)

I suppose it depends on:

- How many patches you receive

- How easy it is/long it takes to distinguish the good from the garbage

You're assuming the other 70-80% is actually good. It's not, there's a whole spectrum between "junk" and "good". A whole bunch of this stuff is "technically correct, technically even a security issue, but also just plainly not of actual importance".

You get free meals, 70-80% of them tasty on different degrees, and the rest burnt. That's OK because free meals, right? Except the small letters say you have to eat 300 meals a day.

I won't claim that my experience is representative, but what I've seen is closer to 50% being false positives, and then maybe 20% of the remainder being bugs but not vulnerabilities (aka lower priority bugfixes). The suggested patches have a much worse hit rate though, to the extent that I accepted only a small fraction after reviewing.

That said, an automated system generating only 30% true positives is probably still ahead on ROI as long as we're talking about tens of findings and not hundreds. The amount of human time spent investigating things that turn out to be false positives is the main difference.

Yeah, I was being "optimistic" in my estimate. And of course with a batch, you start at the high severity end, so it gets worse as you progress through. (Well, I assume at least, even some of the "critical" was just plain "not even a security issue" in our case.)

> If the concern is cost, Anthropic offers a Claude for OSS program.

thanks! It turns out that I am apparently qualifying for it and I was unaware

lets see whether Anthropic agrees

> Anthropic needs to teach their systems to propose a fix too. Detecting a coding problem and proposing a patch is what we should expect from a company with supposedly the best coding LLM.

Of course it's trivial to have an LLM generate a fix, too. They probably do. It doesn't mean a project (1) wants to take an LLM-generated fix at all, or (2) can immediately commit an LLM-generated fix without review (which takes time).

Punchline: Don't use curl.

Maybe not. And that’s totally fine.

But right now they’re not even offering one. They’re just putting the burden on the project without even trying to help fix it.

They have all the resources in the world, and they won’t even try.

And that’s of course assuming it’s a real bug, as we’ve seen a ton of them don’t seem to be. The last time we went through this, curl reported not only that a bunch of them weren’t bugs, but some of the bugs were listed previously on their mailing list, and that’s probably where the LLM got them.

So not only is it finding small bugs. And nonbugs. It’s reporting previously found bugs.

Pilling work on maintainers. Over and over again. To help? No. To make big numbers to show off for their IPO documents.

?

https://news.ycombinator.com/item?id=50448558

Could also be that most of the findings are low signal noise. Even the batch that was sent to curl was one medium and 39 low severity. One can only assume that the vast numbers of findings that haven’t been sent are even lower severity.

I'd be careful to take an examplary projects like curl and assume that level of quality is the average or even close to it.

Compared to curl, even the Linux kernel is garbage, and the Linux kernel is probably higher quality than all of the enterprise software that runs the world.

The most likely explanation is that it isn't worth submitting a massive stack of bullshit to human review.

If they actually were sitting on a PR gold mine of vulnerabilities they wouldn't be throttled on review resources while artificially inflating/misreporting their numbers. They would be diverting resources to clear the review queue.

It's fake and it always was fake.

> Of Anthropic’s 37,966 reported findings, only 7,349 (19.4%) have been made public and just 217 have been marked fixed.

So the article is making a sweeping statement based on the 19.4% of publicly disclosed CVEs?

That statistic suggests the opposite. I'm sure government is aware of many back doors they keep secret and prefer them to remain secret.

"public" here has nothing to do with government holding on to vulns. it's about landing fixes and coordinating complicated / impactful findings

You really think Anthropic would publish stats that say "Yeah, we've found 40k vulns but are going to keep 80% of them to ourselves [and the govt] because those are all the juicy ones"?

Your assumption about governments hoarding zero days is accurate. There's a book on this called "This Is How They Tell Me the World Ends" written by NYT Journalist Nicole Perlroth.

Very interesting that the conclusion of the article is so tightly aligned with the VulnCheck's business model! It's almost as if this is just a thinly veiled advertisement...

It is bad if this points to a general pattern of AI being better at finding than fixing vulnerabilities. The larger that ratio, the harder to secure all the things. The current stat from this experiment is 37,966 to 217. I hope they didn't apply equivalent tokens to both sides, or we're in trouble.

It'll never scew to far of as long as defence is a RL target. Better red team just means the blue team has stronger target environmenta to RL for.

So far the trend is pointing to AI being better at finding problems than designing software to avoid them.

Which I'm fine with really, but it means you need to be very careful letting it iterate on code until it finds no problems.

If anyone here hasn't experienced agents going into a loop finding "problems" and "fixing" them by adding more brittle complexity which introduces 10 more problems I'd be surprised. It's like the biggest experience of working with agents RN

The worst part is being thorough takes so much more work - they often do raise valid problems but you just can't set them in a loop because of this tendancy so you end up with way more work and hard questions

have you tried modularizing / breaking down your code into different files (ideally each having functional code so they don't depend much on anything else) as much as possible?

are you suggesting this because you think I haven't tried it? I'm genuinely confused. We are professional developers working with a team of more than 1 person. Yes we have code in multiple files and have heard of the basic ideas of software architecture.

It's like being asked if i'm using source control? It's kinda a given if you're doing more than hobby projects.

I'm saying lean more into it, to the extreme and instruct your models to do the same

How would that help with the raised issue?

Smaller the function, less ways to screw it up, for humans and llms alike. I suspect humans don't code like this because it is too tedious but LLMs don't have such problems

Less ways to screw up an individual function, more ways to screw up their composition. Still not clear to me that it's better

That's my biggest issue with it personally, as an academic. Way too much unnecessary indirection that makes it harder to read and understand the assumptions and parameters going into the code, and unnecessary defensive programming for code that is just going to be run a few different ways with strictly controlled inputs.

I knew it. I keep asking people why they're constantly checking if an attribute might be empty.

One particularly bad instance consisted of someone checking for the existence of an attribute on a class they had created themselves.

When I asked why they didn't simply ensure the attribute existed they got defensive.

I have. My main 'claim' is that AI isn't naturally inclined to structure code well.

Doesn't mean you can't make it, or rather you must.

I have a "review panel" skill, that will do multiple rounds. But after two rounds, it looks for hot spots. If hot spots are found, it triggers a "big picture" design review.

It's been helpful.

yeah, thats similar to our current approach. It was inspired by a blog post from anthropic where they talked about the same kinda idea, where iirc after 3 rounds the author is asked to rearchitect.

The only other bit we also stole from that was the idea of only looking for issues in the original diff rather than spawning new problems from each diff fixing the previous. Of course the possibility exists of bugs being introduced while fixing others but we try to handle that with human review for the minute because its not worth getting agents to cover that bit, at least with naive approaches

So true, I had to learn this the hard way. I ended up doing a convergence review after each loop iteration and suggesting scope to cut where findings were diverging.

It is my biggest problem right now: AI can only add things. It does not notice that things can be removed or reworked better.

This is true for both code and prose. Even when it does notice something wrong, it wont remove it, but keep the sentence and make it longer.

"This is A" -> "This is B, not A"

It makes it incredibly hard to write plans and documents and maintain code.

When I tell it that the file is getting overlong and complex, unless I mention explicitly each single point, it just plainly doesn't see the problems and instead removes important bits. I did not yet find the right wording to express it to AI in an automated way.

Coding is hard. I wouldn't be surprised if any patch they suggest came with new bugs, as it often happens with humans.

once you have documented test for a bug AI is VERY good at fixing vulnerabilities. If there is a human there to look thru the fix code

Does the author expect that every company using Glasswing/AI is going to credit the AI directly? I would not expect Microsoft or Cisco to credit Anthropic if one of their internal teams uses AI to discover a CVE in their code.

Attribution/reporting is probably part of the conditions for these companies to get access.

In my opinion, what tells a lot about AI provider AGI claims versus reality is the difficulty to have LLM provide the fix to issues.

For PHP2shell vulnerability for example, there was "patched" .deb (and equivalent) packages only for the most recent and most trendy distributions but countless older versions of distributions and php versions were like abandoned without simple solution. I would have guess that "patches" and pre-built fixed packages would be flooding the world. And with just a prompt of one line, most servers would be properly patched. But no.

Which AGI claims are these specifically?

I find it hard to assess when people mention false claims when I can't find what the claims were to determine what about them might be false.

Expected this to happen maintainers overloaded with issues majority of which are nothing burgers.

AI models generating slop findings? Who ever could have predicted this?!

The PHP2Shell vulnerability ended up being paradoxically comforting for me.

They turned frontier models loose on PHP, and the big one they found was based on a previously know logic flaw that was extremely difficult to exploit. Which makes sense: hence the need for frontier models.

And in the wild, while there have been many attempts and a few simple demo successes, the practical impact has seemed to be small, seemingly constrained by the length of the chain and the diversity of PHP configs.

In short, it was indeed a major vulnerability, but not magic (in the Arthur C Clarke sense).

I think ultimately the fear comes down to: do you believe that code can be secured. I recognize that there is probably no objective way to deductively prove it because of the Halting Problem etc. But if one believes code can be constructed to operate to spec, then there must be a limit to the damage an LLM can do, even a very smart one.

I think we are going through a phase where a lot of known triaged code shortcomings will get exposed and hopefully fixed. Sucks while it is happening. But not actually the first round of this type of evolution. And ultimately I think as before, we will come out the other side with more secure and durable technology.

Thus far, these agent vulnerabilities are all demonstrating million-monkey intelligence, rather than "super" intelligence, in the sense of some swarm of robo-Neo super hackers.

That's consistent with my experience using them for coding, where even the best models mostly work by being persistently midwit (with infinite knowledge) more than individually brilliant.

“LLM” may not literally stand for “Lots and Lots of Monkeys”, but this intuition can nevertheless take you pretty far!

I think the more fundamental question than "can code be secured" is something along the lines of "can we properly measure the security of the various components of a sufficiently codebase". Security consultants have workflows but ultimately everything is ad-hoc and suffers from cominatorial explosion, and what's missing is the security equivalent of Lean, a real formalized effort to prove security. There have been many attempts, of course. One of the great problems of our era.

*sufficiently complex

> and hoppefully fixed.

Well, the report you're replying to doesn't make me so hopeful: "7349 findings made public, 217 (.5%) fixed"

The halting problem is a product of a Turing-complete model of computation.

If we want perfectly secure systems, then perhaps we don’t want universal and general purpose ones.

The bar for Turing-complete is very low. The halting problem is to programming as Godel's incompleteness theorem is to metamathematics. Yet no one is seriously suggesting that we downgrade to axiomatic systems incapable of expressing Peano arithmetic, nor that our current axiomatic systems are impractical for proving theorems.

We cannot automatically and completely verify nontrivial semantic properties of arbitrary programs written in a Turing-complete language (Rice's theorem), but as long as we as humans can intuit correctness arguments for our programs, then formal methods will apply to it. Of course, declarative rather than imperative programming semantics tend to result in simpler correctness arguments, sidestepping reliance on global state and convoluted loop invariants.

I just don't think we need turing completeness to solve the problems we try to solve with software. You bring up how humans can reason and prove correctness for our programs, but I say we just force them to do that (making our programs no longer turing complete). I don't think I've ever encountered a problem that required steps that I didn't think I would be able to prove halted.

One can imagine a world in which Rust allows a `decidable` block and code has stricter guarantees within it.

I imagine it would be a nightmare to work with.

Software “engineers” aren’t responsible for their code the same way professional engineers are. Yet, I’d prefer the embedded system for the supermarket freezer were a nightmare to work with, over a society that is a nightmare to live in because of “cyber.”

And maybe we discover that perfectly secure systems are not useful, or fun, or otherwise enjoyable to use.

Maybe the digital equivalent of being locked in a doomsday bunker eating only white rice doesn't produce interesting software experiences.

The expectations of AI models in security really do seem to be magical in a way that, I think, obscures what needs to be done. Of course AI can’t magically create exploits that somehow drill through otherwise secure systems, because computers don’t work that way. It’s just a matter of not taking for granted things that worked in the past, like downloading code from the Internet and running it in a super-duper secure sandbox.

A computer that isn’t connected to anything (make sure to disable Bluetooth and the microphone) is already safe. I bet eventually we’ll be able to serve static HTML pages safely eventually. Maybe not images. Font selections might have to be pretty restricted. But, at least we’ll have the useful part of the internet back.

>A computer that isn’t connected to anything... is useless.

Just get rid of the computer at that point.

> I recognize that there is probably no objective way to deductively prove it because of the Halting Problem etc.

There is. But it will only be reliable if every transitive dependency of your codebase, all the way down to the hardware level, has been formally verified. Failing that, you have to rely on opaque contracts at the boundaries, and those are just promises that may be mistaken.

The Halting problem, Rice's theorem, etc. only apply if your goal is to automatically deduce properties or their violations for arbitrary programs. If your goal is simply to prove correctness properties for well understood classes of programs in industry, then there are no fundamental obstructions.

Formal verification is not magic. It can execute deductive proofs, but the relevance and utility of those results is an question of human values. Spec bugs are a thing. Bad design is a thing. Out-of-scope usage is a thing. Even a halting oracle won't help with those.

We come out stronger. The unpatched long tail is going to have a rough couple of years though.

There was a lot of hype early this year. Mythos, openclaw, gastown. And now they’re basically completely irrelevant. Things move so fast.

[delayed]

Does anyone know how Glasswing relates (quality-wise) to https://github.com/anthropics/oss-scanner ?

It's in the article.

No, it's not, my question is how large the difference in quality is, as a result of the missing validation.

Hard not to draw parallels with Deepmind’s GNoME project which found millions of new “materials” most of which turned out to already exist or be garbage basically.

https://pmc.ncbi.nlm.nih.gov/articles/PMC11044265/

The common thread is asking humans to do the verification work, which is the hard part.

>most of which turned out to already exist

And if you came up with a model that discovered them without previous knowledge of them that's a pretty amazing tool. Huge numbers of materials are sat on by companies and little to nothing is published on them.

As someone working at a bigtech co that has spent every single working day (and many weekends) since the launch of Mythos patching serious, company-ending type of security issues, these types of posts really irk me.

Public datasets of CVEs being underwhelming doesn't mean much if all of the issues are in private company codebases that will never get a CVE number assigned and will never get a fancy blog post. You're never going to hear about them, and the people working on them probably can't tell you about them, but that doesn't mean they don't exist.

We're all worried about OSS having security issues because of how widely used things like PHP or curl are... but the same can be said about email providers, document collaboration SaaSes, payment processors, etc that millions of companies depend. Those companies will never be in datasets like this, and they'll never publicly disclose how many issues they found.

Yeah I am much more worried about some critical vulnerability in, say, the controller logic for pumps at sewage treatment facilities vs. a sandbox break in Chrome (which will be patched well before the blog post...). The broad issue is that you can turn DeepSeek-v6-unhinged loose on an arbitrary IP and exploit ~everything that is not developed and actively maintained by a billion-dollar company

One example they use in the blog post is curl. Not maintained or developed by a billion dollar company.

I suspect open source projects are better off than proprietary software - they already have many eyes on them, whereas commercial software that’s not a prominent and profitable target (e.g. windows, ios) hasn’t had much attention until LLMs made attacking them significantly easier

The average enterprise software is swiss cheese. It seems the hacker news type that likes to post just doesn't realize how much business/enterprise software exists out there. An F100 company has thousands, maybe tens of thousands of applications that you've never heard of and never will unless you work for them. Industry has applications ran by relatively few people, but it's keeping processes that make the modern world work alive.

Worse is everyone is like "Well, this stuff is protected or air gaped". Yea, it's behind firewall, or even private networks commonly, but the moment you get inside that network you have tons of lateral movement and could cause unbelievable damage in minutes.

    > Not maintained or developed by a billion dollar company.

Directly? No. Indirectly? Yes. They are mostly funded by billion dollar companies.

Mostly funded is different from “maintained” or “developed”.

Afaik everyone working on curl has a day job.

> I suspect open source projects are better off than proprietary software - they already have many eyes on them

This has been disproven by the sordid history of OpenSSL. It was open source, critical, and full of holes.

https://www.mend.io/blog/how-the-heartbleed-vulnerability-sh...

Single datapoint does not prove anything at all, but it is a data point.

That infrastructure is air-gapped, or should be. Im not worried.

Just like the European freezers and refrigerators.

we've already had water treatment plants hacked and the amount of chlorine in the water changed. Florida ended up with dangerous amounts of lye being added to the water and was saved by an employee that noticed the change...I am worried, no airgaps to be found..

https://www.controleng.com/throwback-attack-hackers-attempt-...

https://esemag.com/water/russian-hackers-breached-quebec-wat...

https://www.cbsnews.com/news/us-water-systems-cyberattacks-i...

https://www.bbc.com/news/world-us-canada-55989843

edit: formatting urls

> I am worried, no airgaps to be found..

Reminder that Stuxnet managed to jump an air gap (via USB keys):

* https://en.wikipedia.org/wiki/Stuxnet

oh great..I'd almost forgotten about that... what I found really funny was they just infected every USB key that vending machine had until a worker bought one...No fancey targeting, just cheap usb keys and patience.

Have you tasted Florida water, it could probably use some extra chlorine. They dont exactly invest that much in public health/infrastructure, im not surprised this happened in Florida. Its a careless state with careless politicians.

Im still not worried about it. There's thousands of water treatment facilities in the USA and you've surfaced two occasions.

Ya'll should try actually breaking into something semi secure with an LLM, its not as easy as you might think. Seriously go try it, if you think its so easy there's millions of dollars to pick up off the floor on bugcrowd or h1. Seriously, please go give it a go.

Secondly, maybe the USA should stop committing mass murder across the globe if we don't want to be the subject of retaliatory attacks.

I actually 'surfaced' attacks in 5 states, and you only read two of them...not to mention Canada and Israel. Also, you're acting like you think its some script kiddies gonna try this stuff. It will be state actors, and was mentioned upstream in the stuxnet comment.

And you think state actors are only capable of harming you now because of llms??? They don't need llms lol.

I would argue that systems were largely more vulnerable 5 years ago than they are now. Esepcially at a municiple level.

There seems to be a lot of "strategy of tension" being employed in these news cycles surrounding security and AI. Dont fall for it. You're fine. Systems are more secure now than they ever have been in the past.

But on the Israel note, thats awesome I hope more of that comes their way.

I never mentioned LLMS, a bunch of the attacks I listed pre-date LLMS, what makes you think I care about LLMS ?

Maybe learn to read and see the parent comment.

LOL - funny comment coming from you... you said "That infrastructure is air-gapped, or should be. Im not worried." I responded with a number of news stories showing that is false, and its not air gapped. You then started harping on LLMs ??

Please don't respond to a bad comment by breaking the site guidelines yourself. That only makes things worse.

https://news.ycombinator.com/newsguidelines.html

We've banned this account for repeatedly breaking the site guidelines.

https://news.ycombinator.com/newsguidelines.html

My sweet summer child....

Please omit snark and internet tropes on HN. The guidelines make it clear we're trying for something better here. https://news.ycombinator.com/newsguidelines.html

"should be" is doing a lot of work here.

Airgaps don't solve everything either. Stuxnet was successful at destroying air-gapped machines.

Airgaps still should be bare minimum though?

Do you know what a modern water/waste processing facility looks like? Cursed controls on Youtube is a good channel to look at. He deals with the electrical parts of the system, but you can see how many of these systems have ethernet ports and are programmable. They also report back to centralized locations so they know when to send this guy out when something is acting up. Once/If you get in the centralized system you can take everything down.

> The broad issue is that you can turn DeepSeek-v6-unhinged loose on an arbitrary IP and exploit ~everything that is not developed and actively maintained by a billion-dollar company

Agreed. My fairly obscure company got hit by something like this last weekend. What stood out was the speed of the loop: it found something small, used it to uncover something else, and pivoted to clever exploits for the new thing within minutes. It hit a dozen of our apps in parallel across within the span of a day, from a rotating pool of residential proxies, Tor, GCP and Cloudflare IPs.

I'm fairly confident it was LLM-driven. The payloads and throwaway accounts had names like "PT59553 TEST IGNORE" and "PocCorp1659r2", and it adapted to error messages the way a person reading them would, just never sleeping. A non-tech company with 1000x the surface area and less monitoring could be in big trouble.

Of course I used LLMs for the defense too, and they were amazing at log forensics and fixes. The annoying part: when I asked Opus 5.5 for clear repro cases of exactly what the attackers did, I hit the cyber safeguards. It felt like Opus was intentionally trying to keep the exact exploits for my own bespoke applications away from me, though it was happy to point out how to fix it.

the broad issue is actually that those are connected to the internet at all and there is no real enforcement or penalties for it

billion dollar companies are not exempt from this, in fact they tend to be the worst offenders

as a bystander not even working in the sector, your witnessing is exactly what I was worried for. if an agent is wily enough to defeat a security it should also be wily enough to cover its tracks, otherwise what's the point?

[delayed]

Have you seen any patterns for the causes of the error you have seen? For example, more training for developers? Or maybe just systems were really very complex that led to bugs?

I'm sorry you've possibly been wasting your time, but this lends little credence to the idea that what you've been patching has actually mattered from a security perspective.

If there isn't data available to assess impact objectively, then this is just fluff and nobody has any reason to believe you.

I am unaware of any comprehensive analysis demonstrating that LLMs can identify and exploit private code repositories for major enterprises any more than they do open source technologies. Are you?

I don't think it matters. You have to address it or it just continues to generate noise. People aren't great at predicting what could really be exploitable, so who is going to sign-off on ignoring it at a large company? And there is real risk there - even if only a tiny fraction are truly exploitable in a way that would cause real harm you just don't know which are which.

You sound like you’ve already made up your mind. Why would anyone waste their time trying to convince you, random internet stranger? It was just someone sharing their personal experience, which you’ve neglected to do, I might add.

I've only made up my mind to the extend that I'm naturally skeptical of anyone selling anything, and to the extent that I've worked in cybersecurity and know how the sales engine works.

Beyond that, I'm confused. OP gives data supporting the idea that security concerns have been a puff-piece PR stunt. Glasswing's numbers are fake, and beyond that why on earth would a company not divert resources to processing vulnerabilities if they were sitting on a PR goldmine?

The response is "trust me bro, I've been working hard to fix stuff for a private company." Which doesn't tell us anything.

The OP gives us ample reason to believe that this commenter has been wasted their time on non-issues caused by LLM slop-reporting. This is consistent with my professional experience (and I'm confident resonates with the experience of many who actually work in the field).

The commenter theirself just says "trust me bro, private is different from public" which I do not have reason to believe.

I work for a private enterprise that has been deluged with security issues that we are fixing while adding zero value to the company due to corporate's response to "Mythos."

We are finding that almost none of these issues are valid.

I would love - LOVE - to hear some kind of valid, quantified, well-argued justification for the titanic shift of resources away from productivity and towards security.

I would love to look at actual data, or for you to make an actual argument, or for someone to make some actual sense on the cybersecurity doomsday front. I engage in good faith with good faith arguments. I have not heard any.

Possible you might be believing the LLM too much. It seems to flag so many “critical, company ending” security issues, the ones surviving any semblance of scrutiny I’d estimate at 0.5%.

Make sure you’re still thinking.

Yeah, I'll believe you, pulling the numbers out of your butt, instead of the guy actually maintaining the code.

Your condescendence level is over 9000.

I'm replying to a new, green account, immediately raising suspicion. Secondly, we're both random internet strangers. Why believe one person over another based on two easily falsified claims? In this discourse my experience should be as valid as any other.

I would agree with the sentiment of the commenter though. Software bugs in proprietary software are real, but LLMs tend to overstate their criticality, and/or inside knowledge is at times needed to exploit those.

We've sure seen that LLMs can make good assumptions about blackboxes and, unlike humans, are able to try many paths in a short time. However, one can't neglect that doomsday prayers have been present before the dawn of AI, and rarely companies got hacked.

From anecdata, the human is still most often the culprit for breaches, not the "We managed to penetrate that one byte often enough" exploit.

Not only that, their "fixes" tend to break things

Please don't attack others on HN, no matter how right you are or feel you are. https://news.ycombinator.com/newsguidelines.html

> Make sure you’re still thinking.

Please avoid patronizing barbs in comments on HN. I can believe you didn't intend to be mean, but it's easy for these things to be misinterpreted on a text-only message board. https://news.ycombinator.com/newsguidelines.html

You are right of course, for posterity I didn’t intend this as a ‘patronising barb’. It’s more intended as general advice, which readers can agree or disagree with.

Do you have any evidence for the 0.5%.

It's the number from the article in the post.

With code public many of the actively used and maintained open source projects have better code quality than proprietary or big MNC ones. Forget Mythos level models but even smaller models are able to find large number of issues in closed source tools in my example. One reason obviously is the peer review and quality checks by users/hackers/maintainers over the years.

If $company has had enough "company ending” security issues to keep you busy every work day for the past 5 months since mythos release then maybe said company shouldn't be in business to begin with? Clearly they cannot have taken security serious in any way shape or form.

If $typical_software_stacks have had enough "company ending" security issues ... maybe the whole industry have been seriously misguided for decades? Clearly they cannot have taken security serious in any way shape or form.

That's correct, but this won't dismiss the parent commenter's argument. A world wide industry cannot be held accountable, but a for-profit company's executive board can be.

Neither "shutdown the company with bad code" nor "because there is bad code out there AI driven security findings are unworthy" is the conclusion I would come to.

Yeah. But how about fining the shit out of companies with found vulnerabilities and data breaches? I can guarantee you they would all harden security in a day if the fines were costly enough

I'll stand with you on this one. I used to work for a cowboy company and I disliked being the minority that actually cared about stability while the head honcho did lip service to the concepts everyday.

You say that, but you also are a green account and offer no evidence to prove you are who you say you are. Anthropic has a huge marketing budget. I appreciate you chiming in, but I have to take unverified claims with a grain of salt. At least the blog post is referencing a real project (curl) so we have some evidence to discuss.

You must really be irked, signing up just to post this! ;)

Anthropic agent.

> these types of posts really irk me.

They shouldn't. AI is great at some things and underwhelming at others, and it's never predictable if you're going to get the great response or the silly one and when you'll get it and over what topic. And you and the author combined are illustrating one scenario where that jaggedness shows up.

how big's your team? or is it just you and a few other guys working on company ending shit over the weekend?

Also work at a big tech company that has fixed many thousands of triaged/verified security issues found by Mythos which would never show up in CVE reports/blogs.

My understanding is that few have been critical vulnerabilities on their own, but LLMs are now able to piece together 10+ less critical vulnerabilities to create a critical exploit. So, some of the classic triage practices that rely on standalone CVSS scores should be coming into question.

Even if you have just have 1,000 low severity vulnerabilities today that don't add up to something critical, they might add up to something critical when the next model generation arrives, so you need to get ahead of those latent vulnerabilities. The safest approach is to fix everything, which is understandably difficult for OSS.

If that isn't possible, then we need new ways of prioritizing/triaging bugs... which would also make for a more interesting blog post than claiming Mythos is 'underwhelming'.

The blog is misleading because it tries to characterize Project Glasswing's effectiveness through use of a tangentially related data source. A Project Glasswing vulnerability discovery most often occurs when a Glasswing partner points a Mythos-class model at its own, often proprietary, software. The data source (aka "ledger") is only for the Anthropic coordinated vulnerability disclosure program for open source software: in other words, Anthropic itself pointing a Mythos-class model at an OSS package, and then notifying the maintainer afterward that it had done this, and had findings to share. This OSS work is certainly useful, but the number of fixes reflects - in part - limited time availability by volunteer OSS maintainers, and Anthropic sometimes working without knowledge of the OSS project's threat model (the threat model is explicitly requested in the newer oss-scanner effort). Percentage fixed is often much higher for use of Glasswing with proprietary software, e.g., https://blog.cloudflare.com/cyber-frontier-models/ "Our research with Mythos Preview was conducted in a controlled environment against our own code; every vulnerability surfaced through this work was triaged, validated, and remediated where action was needed under Cloudflare's formal vulnerability management process."

> tangentially related data source

The used curl project as source which’s received mythos reports directly. What’s “tangential” about it?

> So far, only 2 of the 442 have been exploited in the wild, just 0.45%.

> almost none are being exploited

Isn't the whole point of glasswing to patch stuff way before it can be exploited? If the patched code is distributed quickly- there's no opportunity (or incentive) for bad actors to craft and distribute an exploit. This reads like that's being spun to convey that these are "real" vulnerabilities.

Exactly what I was thinking ...

    But are these vulnerabilities actually useful to attackers?

I thought the idea is that they are finding them before they are used in the wild -- how would an unfound vulnerability be useful to an attacker?

It’s almost as if the author thinks Glasswing should arming attackers with vulnerabilities.

This question is a rewording of "are these questions actually high impact?", and is being asked because Anthropic said almost everything they found was high impact.

I see at least two more CVEs in that https://github.com/patrickmgarrity/Anthropic-Credited-CVEs repository that were most definitely "exploited in the wild" (https://www.cve.org/CVERecord?id=CVE-2026-93292 and https://www.cve.org/CVERecord?id=CVE-2026-92729) -- by virtue of being the one at my org who had to deal with it at the time. So I'm not sure what that person is basing this on.

I received 9 reports of Fil-C issues from Anthropic

All of them were real bugs. None of them were false positives.

The quality of the reports was exceptional. (My baseline is doing security defense on a browser engine, so my baseline has a large sample count and is probably representative of what experts are capable of.)

The quality of the proposed fixes was alright. I took some of the fixes verbatim. I used most of the fixes as significant inspiration for doing it my way.

The quality of the test cases was exceptional.

Maybe I’m special somehow! But I just want to call out the enormous chasm between what I saw of anthropic’s capability and what TFA says

Linux maintainer is extremely unhappy with the high false positive rate from Mythos vuln: https://m.youtube.com/watch?v=NnV_cWeoo5Q&pp=0gcJCVgCo7VqN5t...

I think someone who likes AI thinks it’s good, someone who doesn’t, doesn’t think it’s anything impressive, and concretely nothing of any significance has happened to the world, certainly nothing like the crowdstrike shutdown so it isn’t world ending like how the labs claim it is.

It’s true that I’m an AI adopter. My psychosis is pretty advanced. I’m running four clankers in parallel right now.

The 9 bugs I got were clear security bugs under the Fil-C definition of memory safety. One reason why my experience could be an outlier is the effort I’ve gone to defining what the security properties of Fil-C are, to the point that there’s a crystal clear falsifiable boundary. So the bugs aren’t a matter of opinion or judgement. That helps clankers and it helps me when screening the bugs.

> concretely nothing of any significance has happened

This is a good point, but “cyber apocalypse dooming is an exaggeration” is a different statement than “bugs found by Anthropic aren’t real”. I agree that the dooming is an exaggeration. I don’t agree that Anthropic’s bugs suck.

Two reasons why bugs-found-by AI-are-really-scary and nothing-is-happening can be true at the same time:

1. Anthropic is sharing the bugs found by their most capable model with defenders, not attackers. Defenders are working overtime fixing those bugs before the embargo runs out.

2. The attackers that have access to Anthropic-class capability are using their bugs surgically. They aren’t using them so widely that folks would notice. That strategy is business as usual for advanced attackers

I obviously don't know if this applies for the Fil-C bugs but what we saw in Rust is that there's plenty of "technically that's not OK" which is very distant from "this will actually cause real world problems" https://forge-blog.github.io/posts/rust-std-memory-safety/

For example it's definitely not OK that you can impl Pattern in a way that causes non-UTF8 data in a string with safe code. But you probably didn't let random attackers impl Pattern, on a nightly Rust where that's even possible, and so that's not going to cause problems in our real world - even though it's wrong and must be fixed.

Most of the bugs were things that fall into the “won’t cause real world problems” category.

But they were all things that unambiguously violated the Fil-C promise

I am speculating on a limb here: maybe the Mythos zero days developed into exploits had a few of the "usually won't cause real world problems" happenstances.

But a model with all the world's knowledge would know how to weave them together and successfully exploit.

I like to think that tomorrow's models are going to be more capable, what we are experiencing now is a fire drill to prepare for that.

> One reason why my experience could be an outlier is the effort I’ve gone to defining what the security properties of Fil-C are, to the point that there’s a crystal clear falsifiable boundary.

Building software boils down to defining things correctly. So, it sounds like you did most of the work here. Imagine that you took your definition and moved it into a formal specification (programming language) and then ran that to verify the security properties of your system. What's different?

Or, imagine that you took your original code as your specification language and then iterated on your understanding of it until it was correct. What's different?

Attention highlights and resolves bugs, so that's the baseline. In order for AI to get value, it has to be some kind of multiplier on your attention, but the reality is generally the opposite.

> Linux maintainer

I wonder what it's like maintaining a project where you regularly get competent contributions. No wonder they are less impressed with AI - they are the 0.1% of maintainers who had something better.

I see it the other way; Linux is a huge project, but being a solo maintainer for a mid-sized project like Fil-C is a different level than one of dozens of maintainers for a large project like Linux. Sort of like being a solo founder for a semi-successful startup is different than being a rank engineer at Google.

> and concretely nothing of any significance has happened to the world

I would hypothesize that all the long-held "easily discovered if you look in the right place, but never reported and so never fixed" 0days in operating systems that criminal gangs and militaries have been relying on (to do cyberattacks with, unlock devices for forensic data recovery with, etc.), are increasingly being exposed and cleaned up, much faster than new such exploits can be discovered; and that this may have given an edge to whichever side of these conflicts don't rely so heavily on hacking the other to achieve their goals.

What do you mean nothing of significance has happened to the world?

You're a marketing plant is what you are

I'm not surprised we see a big difference between extremely mature systems like curl and Linux vs Fil C.

But I'm incredibly concerned by this statement:

>My baseline is doing security defense on a browser engine, so my baseline has a large sample count and is probably representative of what experts are capable of.

This is such a profoundly arrogant and ignorant claim. I have rarely, if ever, seen such sweeping claims from seasoned experts in any field. "Representative" is one of the most notoriously elusive objectives across all of science. Perhaps you have a point, but it's hard to take you serious with nonsense like that.

You broke the site guidelines with this post. Can you please make your substantive points without swipes and avoid crossing into personal attack?

https://news.ycombinator.com/newsguidelines.html

Your response seems a bit overwrought, to call their claim nonsense solely because they used the word "representative" to describe their experience (and they did only say "probably representative"). Their not making a scientific claim, they're reporting an observation about 9 bug reports.

I have these things running 24/7 working on apks. They are extermely underwhelming, I spend most mornings annoyed looking at hallucinated/incomplete/insignificant findings from overnight.

Im about ready to stop using them overnight in this autonomous fashion. Its a great "intelligent" fuzzer but pretty mediocre when its steering the ship by itself.

can you elaborate?

Im not sure why I got downvoted here. Anybody that does this everyday knows you cannot just find exploits with llms without holding their hand and steering them in the right direction.

I cannot say what exactly I work on, but with llms its typical static&dynamic analysis on apks. And they really wouldn't have any sucess without my verbose skills, direction and tooling. They cannot operate by themselves and surface anything significant, atleast not in my experience.

AI is a legitimate step change in every domain of knowledge work, cybersecurity included of course.

OP's post adopts a deliberately skeptical angle, it is a blog post for a company that advertises the ability to spot and patch vulnerabilities. They have an intrinsic motivation to minimize the believed effectiveness of wholesale SOTA-model cyber superiority.

A simple reason could be the age of the codebase. New code contains more serious bugs, and young projects are made up of almost all new code. Mature projects are mostly old code, and many serious issues have already been found and fixed.

It’s not a reason for the wide chasm though.

If you told me that some significant number of Ant findings were duds, then I could reconcile that with what I saw first hand.

But TFA is making a more extreme statement

Likely difference between individual experience and group experience.

Useful to figure out if there are other cases like yours, so then commonalities can be inferred.

Totally

I’m just sharing my data point

The reason is that's just how AI is. It has a very jagged edge of experience. In some ways it's bizarrely wrong and in some ways it seems like a mystical crystal ball.

One thing I should've expected in those reports but didn't is that some maintainers end up nitpicking the reports instead of just doing the best to address the issues. When discussing the LLM code reviews, one author was arguing the security reports are really bad - it turned out that I completely disagree on the judgement they made. Instead of accepting that something can be improved, they argued how unlikely something is, how the project doesn't make this kind of guarantee, how that's an API user error... instead of saying:

- Yeah, we could check the parameter here. Currently we don't return any result so let's add a success/failure response rather than silently failing.

- This is a highly unlikely situation, but let's harden it anyway in case things change and it bites us in the future.

I don't know how many end up in that situation, but it was a whole category of people claiming the LLM reviews are bad, that I haven't really considered before.

Yeah, I don't think we can ignore the small issues anymore.

No matter how unlikely the issues seem to us, an LLM will happily chain dozens of them and do complex timing attacks to eventually gain access.

The post is making a statistical statement, not a universal one, and the examples it gives are of established codebases. If most reviewed code is older and well-used, I think we're seeing what we would expect to see (assuming the models almost always find something to report).

I suspect the gap comes from whether or not you're being used for marketing/IPO bubble driving. I love Fil-C, and its still a small enough project that it doesn't make front page news if a problem with it is discovered. So I think the gap comes in in that they have no incentive to massively drum up the number of security vulnerabilities that they find, by publishing a huge number of false positives that then get quietly retracted

By contrast, when <tool> announces they've found XXXX vulnerabilities in Linux, it makes the news literally everywhere. Even if 0% of them turn out to be real, it doesn't matter - they've already gotten what they want out of it which is the round of coverage about how incredible <tool> is and its the end of the world

That's interesting. While the curl maintainer seems to have been underwhelmed with Mythos 5 in particular, it looks like he also backs up your account of high-quality AI reports (although unclear if these come from Glasswing or elsewhere) at this point.

> The slop is gone

> A sizeable portion of users seem to still believe we still suffer from a lot of made up and hallucinated reports by these AIs. We really don’t.

> Almost all the stupidity and lies we saw in the vulnerability reports made by AIs in the past have ceased. Since spring 2026 the slop reports are practically gone. Our current challenge is much harder: high volume, high quality.

https://daniel.haxx.se/blog/2026/10/22/ignore-the-slop/

>I received 9 reports of Fil-C issues from Anthropic

Probably curated

All of the bugs Claude (Fable or Opus) have noticed as I've been working in Xen have been real bugs; but many of the potential security issues were really obscure combinations of things.

For instance, it noticed that during CPU offlining, a page was freed and then something specific was written to it afterwards. Technically a use-after-free; but to exploit it you'd have to somehow know that the CPU was going to be offlined (not something that normally happens), then try to jigger the page to be allocated by someone such that the later write (also not controlled by you) actually did something you wanted.

Or, if you had a machine with over 5TiB of RAM, but for some reason your control plane was only assigned a single physical CPU, and then an attacker organized memory on boot a certain way, then it could crash the system.

The ones above ended up on the "not a security issue" side of the line, but a handful of ones ended up in the "is a security issue" side. These are all kind of excruciating, because on the one hand they seem so unlikely to even be able to trigger that they're hardly worth thinking about; but on the other hand who knows, maybe there are situations where it could happen.

NB these are just things Claude pointed out to me as I was working with the code, not things Anthropic reported.

It may be a rare combination, but I think the future they're seeing sees this exploited in 100% of cases where it exists. There's not going to be a filtering function on effort if you have an AI attacker, which has perfect knowledge and is perfectly persistent.

The post seems misguided or misdirected in my opinion. Not discounting the fact that the frontier models may find existing bugs, that some are labelled at the incorrect level, even some being a non-bug -- fact remains they are discovering vast numbers of issues. It also isn't the job of Anthropic or OpenAI to fix everyone's bugs, they are doing everyone a courtesy to spend their time identifying them in the first place. The bigger issue appears to be that they are finding more bugs than the developer capacity to can address.

This is literally an ad for a service that helps you prioritize vulns! It is incredible how often an alleged study like this turns out to be an ad for such a service. "Businesses finding ROI hard to demonstrate" -> ad for a product that helps you demonstrate ROI. Etc.

> VulnCheck is helping organizations not just to solve the vulnerability prioritization challenge - we’re working to help equip any product manager, CSIRT/PSIRT or SecOps team and Threat Hunting team to get faster and more accurate with infinite efficiency using VulnCheck solutions.

As someone downstream of the company in the posting, i can confirm the tsunami of AI generated vulnerability reports. And so thr human maintainers are stretched even thinner, and most of the bugs dont matter.

The good thing is the push towards an exploitability rating: If you have a vulnerable piece of software on your system, but it isnt running, it isnt critical.

This is very much inline with Greg Kroah-Hartman the long-time Linux kernel maintainer's view Anthronpic's Mythos audacious claims its cyber capability. Majority of those "findings" were way out of proportion and merely "pattern matching" previously discovered methods of exploitation.

[1] https://news.ycombinator.com/item?id=50342622

[2] https://news.ycombinator.com/item?id=49929391

Pattern matching is a large part of the work. That's why LLM's have been so effective. A human can't match the speed of the loop, which is simple but very labor intensive.

And yet every single company is getting hacked. So some of the findings (presumably non-Glasswing) are clearly quite whelming!

Except... not that many companies are getting hacked. Not more than usual anyway, it feels like.

We don't see many significant big data leaks, or that many DDOS attacks.

Well in cases like these, it’s not the many that matter, but the few. Measuring in aggregate is not conclusive.

> made me wonder whether this was more of an overhyped marketing campaign.

This is the entire premise of LLM hype wave. Or any overhyped tech that tries to force its way into every aspect of life.

My AI Maxxer team mate gets politely ignored now when they put up a bunch of completely random, irrelevant "findings" on my PRs. Probably some model like Fable 85.2.12 can find every nuanced "gap" in my dev-env-only shell script, but I'm not fixing that when there is no production impact and I'm trying to make merge the pipeline to make your containers work.

/Rant

The biggest problem is that AI boosters will of course raise their adhoc evidence and they are fucking loud. The same people who insist there's no need to review code because it doesn't need review. But, fuck. How do they are so sure about that if they don't review, it is just circular thinking over and over again.

Sankey diagrams are ruined for me

Huh? Is VulnCheck talking about itself? I’ve come across many false CVEs associated with VulnCheck over the past few months, including this one: https://github.com/advisories/GHSA-vfj7-8cjw-p6xm

that's hilarious, they made such a big deal of the cyber scare

no moat

I rather like how the folks squabbling about the stats about the report stockpile are ignoring the stats about OSS Scanner. [0] The > 50% false positive rate for Curl agrees with Greg K-H's report of a very high false postive rate from Glasswing/Fable. And -as he mentioned in his talk [1] this year- people refused to pay money for Qualys because its 20% false positive rate was way too high. The only way you're going to get folks to pay for Anthropic's vulnerability scanners is if they're either incapable of understanding the details about its performance or they don't give any fucks about saturating their bug-fixers with garbage reports.

[0] <https://www.vulncheck.com/blog/anthropic-glasswing-november-...>

[1] <https://www.youtube.com/watch?v=NnV_cWeoo5Q>

As AI gets better, it feels more and more like each person gets to hold a bomb in their own hand, but at the same time it feels like having a bomb that can just go off in that hand.

Last week I deployed GLM-5.3-Flash on our on-premise servers, ran it against all repos, and in 6 hours it found 13 vulnerabilities across our products/sites, including 1 full account takeover and an RCE on all our sites, which involved a pretty sophisticated __destruct gadget (all vulns verified to work).

As an experiment, I also pointed it at PHP's source code, and it was able to write PHP code that triggered a use-after-free (or a segfault, if you prefer DoS). I failed to find the vulnerability in the list of CVEs, so maybe it's a 0day, I don't know.

Sure, it's not a frontier model, but it can find many issues that were overlooked by all the commercial security scanners we run regularly as well. The nice thing about this setup is that we can run it 24/7 probing our code/infra, constructing exploits, etc. without having to care about token usage/bills. Plus, I think sending all that sensitive information about your internal infrastructure and the vulns found to Anthropic is a vulnerability in itself.

Thats cool I guess. Now do project Daybreak and compare the two. Also how does it stack against other cots solutions? Like the vibe from this article is I would have just as much luck if I used Nessus.

Another Hacker News highly upvoted topic changing the title to mean something much different than what was upvoted.

The earlier title said something like most anthropic disclosed vulnerabilities were underwhelming (? Can't quite remember).

Also, everytime I notice this it's a change from negative AI sentiment to pro AI sentiment title! Haha

The earlier title was probably the actual title of the article, "After 7-months, Glasswing Overwelms Maintainers with Underwhelming Findings"

Yep! I think you're right