Hosts
About the episode
On Friday, the FAA ordered a nationwide ground stop after communication issues. Then, on Monday, United Airlines announced its own tech outage. That same day, several banks, including JPMorgan and Bank of America, said customers couldn’t withdraw funds because of a software bug. On Tuesday, the NSA announced that its communication systems had been hit by another outage. Then Apple reported difficulties of its own. And before all of this, the FBI issued a Flash warning that cybercriminals and state-aligned hackers were using AI to attack businesses and organizations around the world. Suddenly, the most important question in the news was: Why is everything breaking all at once?
To figure out what’s happening and what we can do about it, Derek welcomes cybersecurity expert Nicole Perlroth and former U.S. deputy chief technology officer Jennifer Pahlka. Is this a coordinated, systemic hack? Is it just engineers rushing to fix vulnerabilities exposed by AI? Is the news media simply more sensitive to outages? Or is something else going on?
Subscribe to our YouTube channel here.
If you have questions, observations, or ideas for future episodes, email us at PlainEnglish@Spotify.com.
In the following excerpt, Derek and Nicole Perlroth explore the problems that led to the many software outages in the past week.
Derek Thompson: Before we get to your analysis of why what happened in the last week happened, I would just like to hear, from your expert perspective, how would you even describe what’s transpired in the last week?
Nicole Perlroth: I think the best way to understand it is something we don’t talk enough about, which is that we have this great digital deficit in this country. We’ve talked a lot about the fiscal deficit.
I always find it very interesting that politicians actually run on the deficit and fiscal responsibility, but we never talk about the digital deficit.
And the digital deficit is just that we have been rolling out flawed code. We’ve been leaving that flawed code in systems. We haven’t been updating it for years.
And we’ve also been baking code into every nook and cranny of our economy, of critical infrastructure, our homes, our cars, our bodies in some cases.
And no one ever stopped to think that we were just adding and adding and adding to this digital deficit that would one day come due. And with AI, it is now coming due in a big way.
And what I mean by that is AI models have proven that they have tremendous capabilities to find vulnerabilities in code and exploit it to break into systems on the other side.
So all of these errors that we rolled out in this code and never addressed are now being found by these powerful models and stitched together in such a way to attack systems. And we are finding 3X the number of vulnerabilities using AI this year than we were in the same period last year.
Attackers are exploiting those vulnerabilities using AI at a much quicker rate than we’ve ever seen. By the time a company learns it has a flaw in its code, rolls out a patch, it turns out that on average, attackers have already been exploiting that flaw for seven days now.
What this means is that while we used to be able to live on this Patch Tuesday cycle—where companies would roll out fixes that were discovered for vulnerabilities in their code, companies would have time to update their systems, and we were all on the software manufacturer’s schedule—now we’re on the attacker’s schedule.
So companies, government agencies, they’re all finding it very urgent to install these patches and software updates much more so than they have in previous years or decades.
The problem is it’s really easy to find a vulnerability now and exploit it. It’s still really hard and time-consuming to properly install a patch. And when you install a patch, you do it in these bespoke environments. No environment is the same. You can’t just push a big red button and everything’s fixed.
Usually, it’s likely that something will break on the other end. And what we have seen over this past week of outage after outage after outage is just that: Companies, government agencies are installing these patches and software updates to fix vulnerabilities that were discovered with AI, and they are breaking things on the other end. So these are not cyberattacks, but the impact is the same.
Thompson: You made two really important points that I want to make sure I understand. Advanced AI has offensive capabilities to discover cyber vulnerabilities and attack, but AI also has defensive capabilities to identify those failures in code and then patch them before the attack happens.
You’re saying that what we’ve been seeing over the last few days is not so much evidence that AI hackers are breaking into systems like United, the FAA, but rather that sophisticated AI tools are finding vulnerabilities in all this critical software, and the patches themselves are coming in so fast and messy that they’re causing their own breakages, a little bit like an own-goal situation for soccer fans out there.
Is that right that the attempt to be properly defensive in the face of likely cyberattacks is what has really been causing these cascading failures in the last few days?
Perlroth: So, almost. I think there’s a few caveats there. One is defensive AI, AI for cybersecurity, for defense, is not quite where we need it to be yet.
Again, it’s much easier to find and exploit a vulnerability using AI than it is to fix that vulnerability using AI. We’re not quite there yet. There’s still an asymmetric advantage for offense.
Number two: To say that attackers aren’t using AI yet, that’s actually not true. Separate from all these outages that we’re seeing, we are seeing a huge uptick in autonomous attacks from all kinds of attackers that are using AI.
And a big red blinking light right now is that we don’t have enough people to help companies that have been hit with these AI attacks recover. There is more than a six-month backlog at Mandiant, which is the first 911 call you usually make when you are breached.
CrowdStrike, they are totally out of bandwidth. They are turning breach victims away. They actually restructured CrowdStrike—they announced it yesterday—so that people at the very top of the organization who haven’t triaged an incident in years are being asked and pulled into incident responses because there’s just so many breaches that are happening right now.
So, two things are true at the same time. AI is absolutely increasing the number of incidents and the velocity of these incidents, and it’s much harder to triage these incidents because when you’re hit with an agent, they don’t follow the same workflow as a human attacker.
We had a pretty good understanding of what the kill chain looks like. You break in, whether it’s through phishing or a zero-day. Then you have some lateral movement as they figure out: Where are the systems of interest? Then you figure out how to exfiltrate or hold hostage data that would be particularly useful.
Sometimes you would catch it in the lateral movement phase. Ideally you catch it at the point of breach, but we had a pretty good sense of how these things worked.
With agents, when they break in, they are doing things that befuddle our puny human minds, so it’s much harder to triage these incidents than it has been. So that is a whole category of things that is absolutely happening right now and is going to get worse and worse and worse.
What happened with these outages that are almost simultaneous—FAA, the bank issues that happened, or the bank disruptions, I guess we should call them; the issue on Sunday or Monday with United Airlines, Monday; the NSANet outage yesterday—those are companies or organizations that are trying to get ahead of this huge tsunami of AI-powered attacks by installing software updates as quickly as they come.
And those software updates have not gone through the proper channels, quality control, testing that they might have in an earlier pre-AI era, and they are breaking things on the other end, which is to be expected.
This excerpt has been edited and condensed.
Host: Derek Thompson
Guests: Nicole Perlroth and Jennifer Pahlka
Producer: Devon Baroldi
Additional Production Support: Ben Glicksman

