Plus: The questions surrounding Anthropic, Mary Peltola’s win, and more…
A string of outages has disrupted major institutions across the United States, prompting people to gauge whether the events are interconnected or share a common root.
On November 12, Microsoft 365—outlook, Teams, and Copilot—experienced an outage that was repaired after 48 hours. (Put forward explanation? A network capacity upgrade.) A day afterward, the Federal Aviation Administration (FAA) caused a rare nationwide ground stop, delaying the vast majority of air travel for over four hours, due to a router upgrade that overloaded air-traffic-control systems and made them dangerously unreliable. Then, earlier this week, a classified internal communications network failed at the National Security Agency (NSA); this followed outages at Chase (roughly 9 a.m. to 2 p.m.), as well as at Bank of America and Wells Fargo. (Alleged reason: Chase attributed it to a software bug in its on-premise hardware stack and insisted “there is no evidence this is cyber-related.”)
The Reason Roundup Newsletter by Liz Wolfe Liz and Reason help you make sense of the day’s news every morning.
Most of this seems largely self-imposed: upgrades, maintenance windows, and software bugs. Yet insiders within the FAA and the NSA have pushed back on whether the official explanations are credible. FAA air-traffic controllers pointed out that routine maintenance typically occurs at night on busy Fridays rather than during daytime on a crowded Friday, and they noted earlier technical problems in the day, well before the ground stop. Questions linger about the NSA interruption as well: Has it been repaired? Was redundancy built into the system that would have allowed staff to continue working? How exposed are we if their networks fail? (The answer could be “not very exposed at all,” which would raise the question of whether we even need an NSA!)
Then the FBI issued a warning earlier this week that “hackers are relying on autonomous AI hacking tools to launch cyberattacks across the U.S. and around the world. The campaigns include a broad assault on ConnectWise customers in the U.S. and on South Korea’s financial sector,” according to TechCrunch. In its latest rapid cybersecurity advisory, the FBI notes that attackers have targeted multiple industries in recent months—financial services, government, technology, defense, telecom and internet services, and law firms—using open-source tools like ARTEX to carry out autonomous-style intrusions.
An Advent calendar, but each day a part of our infrastructure fails.
— Douglas A. Boneparth (@dougboneparth) November 17, 2026
“We need far more analysis of why so much is breaking at once,” notes Nicole Perlroth, formerly a cybersecurity journalist for The New York Times. “I don’t want to overstate things, but what I’m hearing doesn’t contradict the message that most of this stems from quirky updates. Yet taken together, the clues are forming a broader pattern. We are seeing:
- “A surge in severe and critical flaws, with AI accelerating discovery to three times the number of vulnerabilities identified in the same period last year. …
- “The average time to exploit those flaws has fallen into negative seven days. AI lets attackers locate and exploit holes far faster than developers can fix them.
- “Robust AI-enabled offensive tools like ARTEX have fallen into the hands of various state actors—many with questionable governance—along with cybercriminals and digital arsonists, which naturally breeds chaos.
- “This creates immense pressure to shorten patch windows and hasten software updates, which raises the odds of fixes breaking things in unforeseen, cascading ways.
- “The push to “AI everything” implies we’re layering automated AI utilities on top of fragile infrastructure that cannot sustain it.”
I share this view, though I would rate the odds higher that certain government agencies and businesses are misrepresenting the situation. It’s possible we’re witnessing just how fragile our cyber‑infrastructure has become because firms are weaving AI tools into the mix and because bad actors can leverage AI to mount more effective attacks. It’s plausible we were in an unusually calm cyber era before, and this could become the new baseline: frequent, rolling outages that are priced in. It’s also possible these events aren’t happening more often, but our attention has sharpened, and we’re primed for AI‑driven disruption. (I’m skeptical of the latter explanation, but it’s worth mentioning.)
My instinct: incidents like these will become markedly more frequent for an indeterminate stretch, driven by rising volatility across sectors and a boost in AI productivity. Perhaps the two trends will counterbalance each other.
A hopeful takeaway for libertarians: perhaps the state’s power wanes as people lose faith in its reliability. A darker takeaway: when the state withdraws, who steps in to fill the gap—and what sort of world does that create?
Thanksgiving could feel a lot like Planes, Trains and Automobiles. Best of luck.
Scenes from New York: New York City public school enrollment fell by 23,000 this school year compared with the previous year, continuing a decade-long decline. Over the decade, enrollments have dropped by more than 200,000. Some label this a City Without Children. They’re joking about a forthcoming book (to be published early 2028) by yours truly. They also suggest every Reason Roundup reader should pick up a copy to uncover the answers and help fix the country.
QUICK HITS
- A particularly thought-provoking Matt Levine piece at Bloomberg examining scandals, corporate governance structures, and how the Huang affair has unfolded.
- Terrifying:
You should read the Vals report. Users of AI models can be conditioned to target civilians more often by simply telling the models they’re playing a game. https://t.co/vPiQST6bpm https://t.co/z5JTbM2dUp pic.twitter.com/HVx07y389t
— Joe Weisenthal (@TheStalwart) November 17, 2026
- The latest results are in: with this final contest officially decided, Democrat Mary Peltola has captured the Alaska Senate seat from Republican Dan Sullivan, giving Democrats a 56–44 majority in the Senate.
- “Anthropic assumes the AI race is inevitable, so no single actor can prevent superintelligence from emerging. The financial upside of building a machine capable of matching or surpassing human cognition is so large that someone will pursue it,” writes Jasmine Sun for The New York Times. “Second, Anthropic positions itself as the most responsible steward of AI among possible contenders. Publicly and privately, Mr. Amodei tends to reference OpenAI and the Chinese government as competitive boogeymen. These players are seen as aggressive in pursuing powerful AI models and their commercialization, but they would be much less prudent—reckless at best, and authoritarian at worst.” Toward the end of the piece, Sun tackles perhaps the most critical question: “Why is Anthropic confident its own power won’t corrupt?”
- “Most of us have enjoyed a era of peace, prosperity, and above all, stability,” writes Josh Tyrangiel for The Atlantic. “We’ve fallen asleep without worrying whether the traffic signals will work tomorrow. But if my sources are correct, we should brace ourselves forSoftware Beirut—an era where nothing stays intact all the time, yet anything could crumble at any moment, and the essential skill will be not to be surprised.”
- The Flock moment might be ending:
Flock just lost one of its five biggest customers as San Diego City Council voted to end its contract and deactivate its ALPRs!
This is the largest city we’ve seen move away so far.https://t.co/IqWeGlk2c5
— Dan King 🇺🇦 (@Kinger_DC) November 17, 2026