System Shutdowns Are Coming in the Next Year | Joel de la Garza artwork

System Shutdowns Are Coming in the Next Year | Joel de la Garza

MTS

October 24, 2026

Joel de la Garza, partner at Andreessen Horowitz focusing on cybersecurity, explains how AI agents and zero-cost knowledge work are triggering an unprecedented wave of software vulnerabilities.

0:00

Joel de la Garza

What we're gonna see is that people will take systems down, governments will take systems down, critical infrastructure will just be shut off until they can patch it effectively and bring it back safely. And so we will see disruptions, probably not in any critical systems, but like, you know, maybe the DMV has to bring their systems down for a day or two to like upgrade it, and then you can do your registration tomorrow.

0:22

SPEAKER_2

And we're back, we are live with Joel de la Garza, who is a partner at A16Z focusing on cybersecurity. I don't really know what we're going to talk about. It's not, you know, it's kind of a slow week in cyber.

0:35

Joel de la Garza

Yeah, I mean, everybody's getting some phenomenal sleep. It's been pretty good. Not a lot of urgent calls in the middle of the night asking for help. It's pretty awesome.

0:44

SPEAKER_2

So like, what's going on? Explain, you know, explain this to the normies.

0:48

Joel de la Garza

Yeah, of course. And I think it's funny, because I think it's sort of coming back to the original sin of sort of the security profession, or I'd say generally just kind of the role. We have these things, they're called Chief Information Security Officers. These are the people that are supposed to be responsible for information security for your company. You probably know them by name. They're the ones that usually send you phishing emails that try to get it to click on links that you're not supposed to click on and stuff like that. But broadly, like the role of the CISO originally, like when it was originally conceived, the first CISO happened at Citigroup. It happened right after a bunch of Russian hackers got in and stole a ton of money.

The first CISO was really focused on the technical security of the systems. The way it used to be way back then in the late 90s was that systems were trivially secure.

You could literally open up Telnet, connect to a port, and start issuing commands as root. That's just how insecure things were at the beginning. As a result of that, there were just chronic vulnerabilities, right? You remember there used to be websites that tracked all the website defacements. The big thing to do was to go hack the New York Times and change their main website and put your hacker elite sign on there.

There was a whole ecosystem of this stuff, and it was just hacking was happening all the time. There wasn't a whole lot of security. Over the years, Microsoft did their trusted computing initiative. People got very serious about security. We started applying patches. The bar got raised. But over the last 10 years, things had been getting progressively better. But into the security world crept this concept of risk acceptance. And so the CSRS job went from being something where, hey, we need to fix all these vulnerabilities, stop all these problems to being one where we have to accept the risk of all these medium to low severity issues.

And we just have to do that as part of the cost of doing business. This is about doing business. It's not about being secure. And the thinking was largely that, hey, look, to exploit these vulnerabilities, to find these vulnerabilities, it's going to take hundreds of hours of human time, of very, very technical people, to figure out how to hack this stuff. And the thinking evolved to being kind of like the same way that we rate physical safes, right? So like, if you buy a safe, the safe will have a number rating, and that number rating corresponds to how long it actually takes to physically break into that safe. Higher the number, the more hours it takes. And so that was the thinking that permeated security, because it was always believed that no one would ever spend the hours needed to get into these systems. And so what's happened in the last two years, three years with this AI revolution is that the cost of that knowledge, the cost of that knowledge work has gone essentially to zero. And so now it's fundamentally trivial to take advantage of a lot of these vulnerabilities. We're finding a lot of issues in code that had been there for many years, just because people are finally looking, or rather agents and bots are finally looking. And so for the next three to six months, we're just going to be seeing a ton of these vulnerabilities coming out, right? Like we had the PHP one, we have one in BIME, there's one in the Linux kernel. There's like four or five others that are currently embargoed that you're not supposed to talk about. But it's just going to be a wave of these things that keeps coming. And so we're fixing problems, we're finding problems that have been there for a very long time. It's just that now we've unlocked this like really great security intelligence that can go find them for relatively little money.

Read the other 20 minutes of this episode here.

$15 gets you this episode and 99 more. Credits never expire. After checkout, one link brings you back here to read it.

Not what you expected? Full refund within 14 days.

Thousands of transcripts fetched by people building searchable podcast archives

Fetch transcripts with code

The same credits work from a script or an AI agent. The demo key returns a sample episode in full, no card needed:

request
curl -H "x-api-key: pt_demo" \
  https://spoken.md/transcripts/1000651996090

Markdown with the speakers named, for your notes, your knowledge base, or anything that makes HTTP calls.

From $0.10 per transcript. No subscription. Credits never expire. Prices exclude VAT, added at checkout for EU customers. Not what you expected? Email us within 14 days with 20 or fewer credits used and we refund the pack in full.

Using your own key:

request
curl -H "x-api-key: YOUR_KEY" \
  https://spoken.md/transcripts/1000797880552