Thursday, November 19, 2026
Newsletter About
Opinion

Everything Is Breaking

Three theories that explain the internet outages

This article was originally published by The Atlantic and is republished here under license.

You know things are bad when panic-buyers start emptying the shelves at Costco. As you might have noticed, everything is breaking. Over the past week, Outlook has gone down, U.S. air traffic control has gone down, banking systems have gone down, and the National Security Agency’s confidential-telephone system has gone down, and just today, Apple announced that some of its services are down.

There is an air of something nefarious afoot. “If this is all a coincidence,” Senator Josh Hawley of Missouri said yesterday, “it’s like the greatest coincidence in the history of the world.” Others have traded apocalyptic theories: Are rogue AI agents taking over everything? Is this the End Times?

The answers to these questions are, respectively: No, AI agents are not taking over, and isn’t it always the End Times?

Let me explain.

I believe strongly in Hanlon’s razor: Never attribute to malice what is easily explainable by stupidity or incompetence. So even on Friday, when the federal government halted all air travel in the United States just one day after a massive email outage, I was doubtful that these events were connected. But now mass outages and vulnerabilities to critical-infrastructure systems have surfaced day after day, sometimes multiple times a day. Even I have found it hard to shake the feeling that this is more than a coincidence.

[Read: What on earth is going on?]

There are three likely hypotheses to explain these events. The first explanation is benign: All of these breakages and failures are happening because we have old legacy systems that are running on outdated code. Crumbly old computer systems are crumbling.

The second is more nefarious: foreign interference. This is not a crazy notion, especially given the crucial nature of the systems that have gone dark. (A special thought and prayer to the lucky ones who had to rebook a canceled flight on Friday night and couldn’t access their Outlook inbox.) The constant onslaught is also suspicious in light of recent events. U.S-China relations are at a low point after Jensen Huang was pushed out as CEO of Nvidia, in part for having an affair with a woman U.S. intelligence has reportedly identified as a Chinese spy. (China continues to deny that this is the case.) If not China, perhaps another foreign actor—possibly Iran.

The third hypothesis is that third-party actors are using powerful AI tools to find a number of vulnerabilities and bugs in our digital infrastructure. But given that there reportedly have been no demands for ransom and no one bragging and claiming responsibility online, as is typical of hacking groups, this explanation seems unlikely. (Many ransomware attacks go unreported.) Hacks are very hard to pull off, Scott Shapiro, a cybersecurity expert at Yale Law School, told me. One hack after another across multiple sectors of critical infrastructure would be a feat so great that it seemingly could be pulled off only by nation-state actors—even in the age of AI. “What’s hard to understand is how these events could be stacked,” he said. “You would think that one is causing the other, but that does not seem to be the case.”

I spoke with two federal officials, both of whom requested anonymity because of the classified nature of their job. Although not directly involved in the response to the outages, they did not see any reason to blame the incidents on foreign interference. That leaves one explanation: This is all (relatively) benign. Indeed, it turns out that a lot of stuff breaks all the time. Major outages on government sites, cloud services, and a number of low-profile systems have been going down over the past few months, all with mundane software issues brought on by new code slapped haphazardly onto old software.

Moreover—and the government officials couldn’t emphasize this enough—high-level engineers who understand legacy code have left their job at the same time that there’s been an influx of people happy to throw anything they vibe code into the existing system. So when things break, as they are going to do, there aren’t people around who can fix them anymore.

As for the timing with the Nvidia fallout, those I spoke with generally saw it as happenstance. Similarly, a number of people told me that the NSA communication system has been slowly failing over the past year and has had multiple (shorter) outages already. “No one is surprised about this inside the NSA,” an employee who works at the agency told me—an observation that echoes the agency’s own public statements.

The biggest tell is in the solutions to these problems, none of which were particularly complicated or revealed any kind of virus, worm, or other type of hacking. Microsoft blamed the outage on a months-old bug. The Federal Aviation Administration cited problems stemming from a router update in Arkansas. Chase Bank attributed its brief outage to a software bug. “The solutions to these problems are banal, because the problem is banal,” one Defense Department official told me. “You run a patch. It’s fixed. That tells you right there what’s going on.”

As tempting as it is to blame these outages on some Mission: Impossible–style foreign-bad-guy operation, there’s a lesson here: When you hear hoofbeats, think horses, not zebras. But that can be a hard sell in the current moment of AI hype. It’s in vogue, among a certain set, to be an AI doomer and to worry about a grand catastrophic event that will render humanity either irrelevant or nonexistent.

There is some truth to this: AI is responsible for the failures, but not because the singularity is here. Rather, most of the complex systems we run on are held together with duct tape. AI itself isn’t breaking those things; we’re using AI to break them faster. Most of our essential systems run on software that is old, out-of-date, and near-impossible to cohesively update. But now code changes are happening so quickly at the same time, and all sectors are losing the senior talent who are in the best position to stop these mistakes in the first place or fix them after they happen.

In that sense, the outages are a preview of what AI disruption will really look like: not some swarm of Terminators, but everyday inconveniences that affect millions of lives and cost billions to fix.

More in Opinion

View All →

Leave a Reply