# AI-Driven Cyber Threats and Supply Chain Defense Strategies

**Podcast:** AI + a16z
**Published:** 2026-08-07

## Transcript

Models are actively escaping their cages, going out on the internet, and doing pretty nasty things.
Recently, we found an API key that had been leaked on the internet that had administrative access to the Apache Foundation.
The interesting thing about cybersecurity in particular is the reward function is incredibly well-defined.
Get access to the data.
Did it get access to the data?
Reward the thing.
For a long time, people had talked about this concept of an NPM worm, this idea that if someone could backdoor a package...
get developers to install that, and then you could use the access stolen from those developers as they install it to self-propagate the worm.
If the labs are making it fundamentally easier to break into supply chain, do you think the labs have a moral obligation to fund some of the problems that they're causing?
I think it's really strange that they're not letting blue teams get access to these tools, but...
AI models are no longer just identifying software vulnerabilities.
They're beginning to exploit them.
In this episode, Joel De La Garza sits down with Dylan Airy of Trumple Security and Feras Abukadije of Socket to unpack what recent AI security incidents reveal about the next generation of cyber threats.
They discuss why frontier models are increasingly capable of exploiting software vulnerabilities, how software supply chains have become one of the weakest links in modern security, and what organizations need to do to defend themselves in an AI-first world.
Thank you so much for joining us.
We've got Farras and Dylan here from Truffle and Socket.
It's great to have you guys on.
This has been probably one of the most interesting weeks, if not the most interesting week in cybersecurity, not because of the Black Hat Conference, which is usually the cause, but because we've now seen several instances where models from not just one provider are actively escaping their cages, going out on the internet and doing pretty nasty things.
And I think Dylan...
Three months ago, I remember a blog post we lightly collaborated on together, and you had found a number of these issues with earlier models, right, that were less sophisticated.
Yeah, we looked at Opus 4.6 and some of the other frontier models at the time.
Given the models a very simple task, there was a barrier which prevented the model from accomplishing the task unless it went and committed a felony and hacked into a system to accomplish the task, but it wasn't instructed to do so.
We found more often than not...
it would do the SQL injection, it would commit the felony, and it would do what it needed to do to accomplish the task.
I think when it comes to alignment issues, no one needs to worry about these models making it materially easy to build nuclear weapons because you need to procure fissile material to do that.
It's not going to make it easier to build weapons.
Everyone needs to worry about these models making it materially easier to hack into things.
The bar previously was just subject matter expertise.
And now the models have the subject matter expertise.
They were specifically trained to have the subject matter expertise.
And they're just making it materially easier to hack into just about anything that you can think of using the fundamentals that we've been talking about for years, but previously required a subject matter expert.
to risk going to jail for hacking things.
DEF CON was always famous for people, for attendees getting arrested at the conference, right?
That's absolutely right.
But that was, I mean, that was a barrier, right?
For better or worse, that prevented these subject matter experts from hacking into things because they were worried about being prosecuted.
The bar has now fallen to just asking the model, which has specifically been trained to hack into things, to hack into things.
So that's a concern.
And then the other concern is when they're incredibly goal-oriented to accomplish tasks, and one of the tools at their disposal...
is cybersecurity expertise.
They will do the path of least resistance to accomplish the task, and that includes drawing on their cybersecurity expertise.
Well, and it seems like, and the classic saying is that don't pick the lock if the door is open, right?
I think that's from the very beginning of the security world.
So it's always been sort of like to go in level of difficulty from easiest to most difficult.
And it seemed like initially these tools had a very finite scope of techniques that they would use, and it seems like they've expanded.
And I think with this test, For us, it was interesting because they now seem to have escaped from just doing things like SQL injection to actually trying to take over packages and do social engineering.
Yeah, it's really interesting to see how, just like humans, the models are easiest path into a company.
And I think that that now has become the software supply chain.
And so just like human hacker would, they're going to pick the easiest way in and the lowest hanging fruit now has become just publishing malware to public registries because they know that there's no vetting happening and developers are likely to install them.
I thought it was pretty interesting.
There was research published recently about what they're calling kind of like universal typosquats or universal hallucinations where all the frontier models all make the same mistake and assume there are certain packages that exist that don't, despite those models coming from different companies.
I think there's just the hanging fruit of a supply chain has just become kind of so appetizing that even the models are trying to get in on the action.
And I think the AI is not only kind of attacking, but it's also kind of the way in a lot of times on the kind of developer side because we see so many.
even non-developers using these tools to inadvertently write code or code comes in, packages come in order to kind of build graphs or visualizations or different things that folks are doing with these tools.
And it feels like no one really knows what's being installed and what's going on.
And this is just basic stuff.
This isn't like, I mean, it sounds like it's sci-fi stuff, but it's really just basics.
What software are we using?
How are we fitting it?
just the basics of computer security.
Can I touch on the supply chain a little bit?
So recently we found an API key that had been leaked on the internet that had administrative access to the Apache Foundation.
And if you're in the shoes of the model and your goal is to get access to some data, certainly backdooring Apache is a pretty effective way to do it.
And to get access to Apache, are you going to use the secret that just allows you to directly log in?
Or are you going to burn tokens and tokens and tokens on trying to find a zero day?
They're optimized to use the path of least tokens to accomplish their goals.
Of course, they're just going to use the secret that's laying out there in the open to accomplish what they need to accomplish.
And so, yeah, I think supply chain and secrets are and have been the path of least resistance and will continue to be so as the models are incentivized to use fewer and fewer tokens to accomplish their goals.
Well, and I think that's absolutely right.
And I think it's that sort of chain of escalation, right, where if one thing fails, try another.
And at the top of that pyramid, right, the top of the hacker ecosystem is the zero-day vulnerability, right?
That's basically finding a vulnerability that can be exploited in a product that everyone uses that you can use to basically unlock all the corporations.
And one of the really fascinating things about the breach disclosure that was made was that there's an incredibly popular CICD tool that I think every enterprise uses that this thing just spat out a zero-day for, right?
And that's just such a critical point in the supply chain that everyone should be thinking about.
Kind of, how are you thinking about that?
That's really difficult.
Like the zero-day creation piece?
Yeah, yeah.
For specific parts of that, like the control of the supply chain.
Yeah.
Well, I mean, the whole world is built on this teetering infrastructure that everyone is using.
It's like the classic picture of the matchstick holding up the complicated machine.
That image probably popped into all the lines right now.
And so everything from package manager registries, like we like to focus on that because what we do at Socket, a lot of those are run by volunteers.
They're under-resourced, underfunded.
Lots of risk there, right?
And that kind of cascades throughout the whole rest of the ecosystem.
So if you look at just the packages that we all depend on, a lot of those are single individuals that there's almost certainly, we know there's a lot of vulnerabilities in the software and there are resources to look for them.
And so I think that stuff's always been there.
It's just that these tools are helping find them a lot easier.
that the frontier models are causing a massive reduction in the time between vulnerability discovery and vulnerability exploitation.
And so what we need to start thinking about is how do we patch more quickly?
And so we need to get away from our take is we can't be requiring our security teams and our developers to do these kind of onerous patch processes where they have to go from some ancient version.
of a package that their team is using up to the latest version across many major version upgrades because that's just so much work to do.
It might require code refactors in your application.
And so if we want to be able to keep up with, a volume is announced this morning and then an exploit is available that afternoon.
We can't be requiring that much work from engineering teams.
And there's also tons of legacy applications that are basically in maintenance mode or unmaintained or don't have engineers that are even assigned to work on them at a lot of the companies that we work with.
And so it's just...
We're going to have to think of new things as an industry for how we're going to patch these things quickly, right?
That's kind of like where our head is at and where I'm spending a lot of time thinking about that lately.
Yeah, yeah.
And I think that goes back to sort of the pyramid question.
So, and I think Dylan, you were hinting at this, which is these are very specialized.
Like zero-day exploits, writing a zero-day exploit, doing a supply chain attack.
Like these are not things that came from nowhere.
These are learned behaviors, right?
And I think what I think we're seeing is we're seeing a...
process that looks like it's been kind of maybe trained or there's a reward structure that's been built on a bunch of these things.
What's your understanding of how they're figuring this stuff out?
Because it seems like they know what they're doing.
Like, they've been taught to do this.
Yeah, I mean, if a lab tells you that this is an emergent superintelligence behavior, they're just lying to you.
And you can read their own safety reports to see exactly how the models are trained and exactly how they're testing these behaviors.
I mean...
The interesting thing about cybersecurity in particular is the reward function is incredibly well-defined.
Get access to the data.
Did it get access to the data?
Reward the thing.
And so when they realized that, like the number of problems that have that well-defined reward structure basically defines how we do reinforcement learning.
And they want to find as many problem spaces that they could do reinforcement learning on.
And so it was a prime candidate for them to come in and give it CTFs.
and give it cybersecurity challenges where they say, okay, get access to this thing and do whatever hacking you need to do to accomplish the goal.
They've essentially been buying pen testing data for the last four years.
That's a piece of it.
The other piece of it is...
And then the capture the flag contest and all those sorts of things.
It's just not difficult to construct a challenge.
Even if there is no known exploit, if we're talking about zero days, you put a piece of software...
between the model and some data.
And you say, get access to the data.
And then if it gets access to the data, you reward it.
And it's that simple.
But the other piece that they've layered on top, and this is where it starts to get really interesting, is they've started to reward the path of these tokens.
And so the reason that's interesting is because for the first time, it's actually able to quantifiably show us.
the path of least resistance for just general cybersecurity to get from A to B.
And we've talked about our opinions of what that is in the past.
Of course, Truffle Security's biased view is a password laying around is a shorter path than going through a fancy zero day.
But actually watching the model physically get from A to B and watching it follow the password and quantifying how many tokens it took to go this route versus that route, I mean, it's just incredible to watch that layout.
And it's all in their safety reports.
Like as they...
test the models out and show, okay, well, it got access to the data and it broke out of its harness, it's not like this is an emergent behavior.
It's specifically trained for this stuff.
It's perfectly logical, right?
Like the fastest way to get a gallon of milk is to steal it.
That's exactly right.
So, I mean, what was interesting is we were in the middle of partnering with Hugging Face to clean up all of the credentials that had been exposed through all of their training sets.
Not Hugging Face's training, but people who hosted training sets.
on Hugging Face.
They use TruffleHug for a wide range of reasons.
And Hugging Face has been a great partner in getting credentials cleaned up.
We targeted their training sets because we knew they had a lot of keys.
Turned out there were about a quarter million live keys in their training sets, many of which had direct supply chain implications.
There was a foundational Linux library that one of the keys had direct push access to.
It could have pushed malware to most machines on the planet.
And so while we were in the middle of doing that, the CTO of Hugging Face shoots me a note and says, hey, This is crazy, but there's this open AI thing that just happened, and I want you to take a look at it.
And sure enough, the first thing listed out in the incident response, although it's true it did utilize zero days, but the first thing listed out was stolen credentials.
And that's how they were trained.
Pathways resist.
Password is a password is always the first step, right?
That's exactly right.
And you've had your hair on fire, I think, pretty substantially for the last 18 months.
I think right now as we're recording this, there's currently an ongoing active breach of a big NPM repo.
Isn't there something happening?
It's more than just a repo.
It's actually about, you know.
A few hundred repos.
Oh, wow.
Okay.
So it's a worm.
Yeah.
And this is one of the things that has been kind of an unfortunate innovation in the malware landscape on NPM is that for a long time, people had talked about this concept of an NPM worm.
This idea that if someone could backdoor a package and then get developers to install that and then you could use the access stolen from those developers as they install it.
to self-propagate the worm, you could create something that quickly takes over NPM.
And this was kind of passed around in blog posts over the years, and no one actually kind of thought to do it until someone thought to do it.
Until someone thought to do it.
Probably using AI, right?
Almost certainly, yes.
And there's been, you know, that malware, I think we have pretty good reason to believe that it was Vibe Coded.
There's been, one of the threat groups actually kind of posted their open-sourced their kind of Vibe-coded toolkit for others to use to be able to do this.
We've seen copycat attacks happen since then.
And malware authors were never really great coders.
You probably realize this, right?
So if the code starts looking better, it's probably Vibe-coded, right?
It's sort of the opposite of what you think of Vibe-coding.
Yeah, and they're using, they're often using the AI tools that are installed on developer systems to go further and to sort of sneak through kind of some of the traditional security tooling.
So we see, you know, basically your local CLI tool often being used, roped into the attack and used as a jumping off point.
So a lot of times the payloads are actually prompts.
That bypasses a lot of typical kind of EDR tooling because it's just like a markdown file that your cloud is running.
It might be given a prompt to search through the system and find all the different keys and things that look valuable.
I'm sure your EDR tool doesn't know anything about this JSON blob in an MD file, right?
Right, and developer machines typically are doing all kinds of weird things.
You are prompting cloud and it's doing a bunch of stuff on your file system all the time, and so nothing really looks that out of the ordinary.
But yeah, so anyway, about the attack this morning, it's a couple hundred packages, a worm spread.
There was a maintainer who, honestly, he's still kind of trying to understand what happened.
We jumped on the phone with him this morning to try to offer assistance.
Oh, that's great.
Awesome.
About half of our team at Socket are maintainers, half the engineering team.
And so, you know, we have a lot of connections in the community.
And our CTO was the former CTO of NPM.
So we sort of try to offer help whenever we can to the maintainers because I do really feel for them.
They often don't know what happened.
I mean, it sounds like in this case, it might have been.
and this is speculation still early, we're still working to figure it out, but it might have been a get of action that was insecure and that an attacker was able to use to ultimately get their code to run and then pull out the token from the environment.
We're still confirming this, but that's what it looks like.
We've seen this with a lot of these worms.
It's likely his endpoint was actually fine and not part of the problem.
It was really just this insecure get of action.
Can I ask a follow-up to that?
So one of the things that I've seen with the NPM worms in particular is after they get their post-install hook and they infect the system, they'll immediately start looking for credentials.
And once they find a credential, they get access to the next system.
So all the time we get questions saying, how do we get the credentials cleaned off the endpoints?
And usually what I say is, well, I'll show you how to scan your downloads directory.
I'll show you how to scan your documents directory.
But in your home directory, there's a place where NPM intentionally writes a credential.
There's a place where Amazon intentionally writes a credential, and that's just how those tools work.
And there's nothing that I can really do to get them cleaned up, even if you move them over to HashiCorp Vault or OnePass.
The credential to HashiCorp Vault and OnePass are on the endpoint.
So I'm kind of curious about that second phase beyond the stopping the spread, but the post-exploitation step.
Do you have any recommendations on how to sort of limit that blast radius or prevent that?
stealing your credentials and jumping onto the next system.
Yeah, I mean, so there's been some changes, some positive movement in the community and the ecosystem.
So one thing that is positive, and it hasn't shipped yet, but NPM has announced that they are planning to, I think it's in January 2027, going to require human interactive confirmation through 2FA before new publishers can happen.
So that will likely kind of kill this whole worm concept completely.
It's going to be super disruptive because everybody's hooked up this stuff to automation so that GitHub Actions kicks off the publish.
And so that's going to break basically the whole ecosystem when they do this.
But I think it's the right call.
But there's other ecosystems that are volunteer run that don't have the backing of GitHub and Microsoft behind them that are going to probably not make those changes.
And so I think we're still going to see stuff like this.
But yeah, we shouldn't have files in our home folders that have tokens and that are long-lived and that let you...
Especially if you're a maintainer with that kind of access.
It reminds me of a friend of mine.
He's a prolific NPM maintainer.
This was back 10, 15 years ago when I was doing this kind of stuff full-time.
I saw him type in his password.
I didn't see the password, but I saw it was far too short.
He typed it in far too quickly.
I called him out, and I'm like, why is your password like six letters, man?
He said, well, you know, He lives in Denmark, which is a very high-trust society.
His worldview about it was that, well, I don't want to live in fear and think about these things.
And I'm like, you're on the internet, man.
People are going to figure this six-letter password out pretty quickly.
And there's a lot of things like that where the folks that are the top maintainers in the world don't necessarily have the security training or even thinking about these things.
And they don't have a security team.
They don't have enterprise SLAs.
These are volunteers that are just putting code on GitHub.
So it's on actually the users, I think, to actually vet what they're using.
It's kind of hard to say, like, you know, we're a company, we just found this code on the internet, and we just deployed it straight into prod, and it's someone else's fault.
And it's like, no, actually, you know, there's definitely some responsibility for the users of this software to really be vetting the artifacts that they're bringing into their environments.
And so I think there's a lot of pieces here, and I wouldn't want to put too much blame on people because it's a hard problem.
But yeah, I think there's a lot of places where we can do good.
Well, let me ask a follow-up to that, because you said there are certain package managers that have resources that other package managers don't.
I think one direct example of this, and I don't cast any blame on them whatsoever, they were actually great to work with.
we found a caching issue in RubyGem that allowed us to steal arbitrary tokens and get access to arbitrary accounts, which we could use to backdoor arbitrary packages.
We disclosed it to them, they got it fixed quick, but that's an example of an organization that's under-resourced.
I mean, how do you think about an organization that's thinking about what technology stack to use and they're picking between Ruby or NPM?
Do you think about that resourcing in that decision of maybe you should consider the one that's got the security team behind it?
It's hard to get people to switch their language completely because of if they're already built on it.
So I don't know if that's super practical.
I think the most actionable thing is folks should fund this stuff.
It doesn't take much money to hire an additional one, two, three, four, five security people that can make a really big difference in these foundations.
So I suggest people sponsor the software that they're using, especially the registries.
It doesn't take very many companies.
throwing in 25K, 50K checks to really make a big difference for these folks.
Or give them some more inference to run the test.
It's funny, I think you guys as being security practitioners, you've had a number of calls with people that want to talk to you about how can I secure my company, how can I secure my business, how can I secure my product.
And those conversations are always some form of I don't want to hire people or pay money for this and how do I do it cheaply.
Thank you guys for coming out to the desert.
You're not here to see us, obviously.
We're all here for Black Hat.
I know that this is neither of your first times here, certainly not my first time here.
I would maybe love to get your guys' takes on what you're seeing.
What's the mood like?
What's interesting?
What are the observations you've got that are catching your eyes?
I mean, at least for us at Socket, I think the biggest thing we're seeing is that 2026 is the year of the software supply chain.
That you're dealing with an incident right now as the conference is happening.
Yes.
And I noticed the attackers seem to pick RSA and Black Hat as the times they want to start these NPM worms.
The hacker, the security designer of the office, yeah.
Yeah, absolutely.
So I think that's the thing that's the biggest, you know, I think.
Prior years, I was having to educate people.
We were always educating people about this problem and having to explain to them, this is not a theoretical risk.
This can happen.
And we'd sometimes get these reactions like, oh yeah, but how likely is it really?
And we're like, no, it's actually very likely.
Let me tell you how it could happen.
And there were many incidents to point to, but I think this year it's really broken through into the mainstream.
And there's mainstream publications.
the business press covering these attacks, right?
It's like throwing a page on Bloomberg.
Yeah, exactly.
So I think that is very, very good because you need that type of air cover for security teams to actually prioritize and find budget for these problems.
And so I think despite all these attacks being very painful to deal with right now, I think in the end we're going to come out really strong from this because we're actually going to get budget and we're going to do a lot of good this year in terms of solving.
It's inoculation, for sure.
How about you?
Yeah, I mean, look, we found a database credential recently that had access to 3.6% of the global PII.
Like 3.6% of the world's population had their PII in this database.
And the reason we found it is because we're partnering with organizations like Hugging Face to just get all the live credentials revoked and cleaned up better otherwise on their platforms.
So it's great to see those types of organizations leaning in.
And like all of the data that's being scraped from all these places, it creates really a central place where we can scan and make material progress.
And so I think like when it comes to secrets, we're kind of seeing the older generation, like HashiCorp got to come.
acquired, CyberArk got acquired, kind of get pushed out, which has sort of led to a new conversation about non-human identity and secrets about how can we make them easier to use, how can we get them revoked, and how can we really move the needle forward with the industry that we weren't able to do before under sort of the old guard.
Yeah, absolutely.
And I can imagine that there's going to be a multiplication effect, which is it used to be one user with 10 passwords, and now you're going to have 10 agents with 10 passwords, right?
I mean, the way agents interact with secrets right now is a Wild West unsolved problem that we're working very hard to solve.
One of many issues.
Thank you, gentlemen, for joining me.
This has been awesome.
It's great to catch up.
Thanks, Joel.
Awesome.
Leave us a rating or review and share it with your friends and family.
For more episodes, go to YouTube, Apple Podcasts, and Spotify.
Follow us on X at A16Z and subscribe to our Substack at a16z.substack.com.
Thanks again for listening and I'll see you in the next episode.
As a reminder, the content here is for informational purposes only.
It should not be taken as legal business, tax, or investment advice or be used to evaluate any investment or security.
and is not directed at any investors or potential investors in any A16Z fund.
Please note that A16Z and its affiliates may also maintain investments in the companies discussed in this podcast.
For more details, including a link to our investments, please see a16z.com forward slash disclosures.