Back to The CQM Brain DumpAI & Automation

Ten Outages in One Week, Zero Confirmed Hacks: What Broke and What It Means for You

Cesar Q.November 18, 20269 min read
Ten Outages in One Week, Zero Confirmed Hacks: What Broke and What It Means for You

What took down Microsoft, the FAA, Chase and Apple in one week?

Short answer: not hackers, as far as anyone can show. Between November 12 and 18, 2026, ten big services went down or slowed down, from Microsoft 365 and the FAA to Chase, Apple and the NSA. Not one has been confirmed as a cyberattack. Of the six that named a cause, five trace back to a change the company made to its own systems. And there is a real AI connection, just not the one you're picturing.

If your Outlook died or Apple Pay refused your card that week, you weren't imagining things. Our phones lit up with the same three questions from clients: Was it hackers? Did AI break it? Are they hiding something?

So we did what we do before any campaign: the homework. We read the incident reports and official statements, checked the numbers against each other, and threw out the ones that didn't hold up. (We also read a lot of Reddit. We regret nothing.)

The week in brief

The ten, in order, with the cause each company gave (or didn't):

  • Microsoft 365, November 12 to 13. Outlook and Exchange Online were down for about 24 hours. Microsoft's preliminary report blames a bug in a July update to its network gateway servers, and says it found no evidence of malicious activity.
  • The FAA, November 13. A nationwide ground stop lasting 4 hours 21 minutes. The FAA says a newly installed router at a site in El Dorado, Arkansas, flooded its legacy network during a contractor's copper-to-fiber upgrade, hitting 17 of its 22 en route centers. About 3,500 planes already in the air landed safely. No indication of a cyberattack, says the FAA, but the root cause (a bad router or a bad install) is still under investigation.
  • Workday, November 13 to 14. During a test, Workday's disaster-recovery systems unexpectedly connected to its live production databases and left the data out of sync. One college documented about 17 hours without it. Workday says it was not a cybersecurity incident.
  • AWS, November 15. About three hours of sign-in and connection failures in its us-east-1 region. No cause disclosed.
  • Chase, November 16. A software bug in on-premise hardware slowed transactions, and payments to other banks and through Zelle slowed with it. Chase says there's no evidence it was cyber-related.
  • United Airlines, November 16. Flight-crew apps went down. United says about 30 minutes and fewer than 150 flights directly affected; AP described a disruption of more than 90 minutes. No cause disclosed, and no indication of a cyber issue.
  • Apple, November 17. iCloud sign-in, the App Store, Apple Pay and more, from late morning until about 5:46 p.m. Eastern. Apple's explanation so far is one sentence: "an internal update caused some outages." Apple says it was not a cyberattack.
  • The NSA, November 17. Its internal network degraded during scheduled maintenance. The Director of National Intelligence tied it to patching newly discovered flaws in PHP, a programming language that runs a big chunk of the web.
  • Cloudflare, November 17. A string of separate regional and product incidents, lasting anywhere from 40 minutes to about 12 hours. Not one worldwide outage, and no root causes published.
  • ElevenLabs, November 18. The voice AI provider reported increased latency and API errors, later affecting some agent conversations. No cause stated yet.

The tally: zero confirmed attacks. Five were changes gone wrong: two updates (Microsoft, Apple), an equipment upgrade (the FAA), a disaster-recovery test (Workday) and a round of maintenance (the NSA). One plain bug (Chase). Four with no cause disclosed yet.

Was it hackers?

We get why this is everyone's first guess. Ten systems fall over in seven days and it feels coordinated. And attacks are very real right now. A cyberattack hit grocery giant Ahold Delhaize in October, and the company expects it to cost 200 to 250 million euros in fourth-quarter sales. Cloudflare fended off 18.2 million network-layer DDoS attacks in the third quarter of 2026 alone, up 208% from a year earlier.

But look at who those attacks hit: mostly mid-size companies, vendors and industrial systems, with the victim usually confirming it within days. The ten outages above don't look like that. And the DDoS flood mostly bounces off the big providers: 96% of those attacks ended in under ten minutes, and none of the big outages in the news has been blamed on one.

Microsoft's outage shows what actually happened. The bug that took down Outlook for a day shipped in July. It sat there for months, doing nothing. On November 12, Teams failed over and traffic went up while about 14% of the gateway servers were already offline for an unrelated reason, and the bug finally woke up. A Microsoft executive said it spread across the whole fleet in under two hours, and at the worst point more than 80% of those servers stopped responding. Microsoft's lab tests had run on weaker hardware than the real thing, which is how the bug survived testing. No hacker required.

Officials agree, for what it's worth. The Director of National Intelligence said the string of outages was not a cyberattack and not a coordinated effort by another country. The Secretary of State said there was "zero evidence" the incidents were linked or caused by a malign actor. Security analysts at Palo Alto Networks, quoted by the BBC, saw no evidence connecting the events, only systems that are tightly connected and fragile.

The FAA is the one to watch: no root cause yet, and the FBI joined the testing. So the accurate phrase for the whole week is "not attributed to an attack so far." Less fun to say at dinner, but it holds up.

Did AI break it?

This one has a better origin story. AI really is writing a lot of code now. Google says about 75% of its new code is generated by AI and approved by engineers, up from about half in fall 2025. The Homeland Security secretary even passed along a suggestion that bringing AI into existing infrastructure might have caused a malfunction.

But no source we found shows that AI-written code caused any of these outages. Microsoft said its outage was not related to cyber or AI. The FAA's administrator said there was zero evidence of anything AI-related and called the AI theory baseless.

The real link is less dramatic and more interesting. AI is getting very good at finding holes in software. October brought a cluster of severe flaws in PHP and Node.js, and US agencies urged organizations to patch, or take exposed PHP systems offline if they couldn't. OpenAI said a security firm found one of those PHP flaws using its AI model. Google's threat intelligence team reported that disclosures of flaws and exploits nearly doubled between January and August 2026. That doesn't prove AI drove the whole rise, but it fits what security teams describe: more holes, found faster.

So companies everywhere are rushing out emergency patches and upgrades, and changes are exactly what break big systems. In one company's tracker of 63 major cloud outages since 2020, about 54% started with a change, a bug or an automation fault. Put those together and AI is adding pressure through the rush to fix things, without writing the bad code.

The NSA's outage is the one officially tied to that rush: the Director of National Intelligence said patches for newly found PHP flaws left some systems unable to restart. That's one outage, on the record. For the other nine, the patch rush is a plausible explanation nobody has proven.

So no, the robots didn't do it. They just made everyone's to-do list longer.

Are they hiding something?

Fair question, because the rules leave room for vague early statements. A public company has four business days to disclose a material cyber incident, and the clock starts when the company decides the incident is material, which can be well after it finds it. "Technical issue" has also turned out to be wrong before: in May 2026, Foxconn first called an incident a technical issue, then confirmed a cyberattack a day after a ransomware gang listed it.

But the cases that flip tend to look the same. Systems stop cold, data gets stolen, and the truth comes out within days. This week looked different. The companies gave specific, checkable causes: a July bug, a router installed during an upgrade, a disaster-recovery system plugged into production, an internal update. That's a lot harder to fake than "no evidence of malicious activity." And keeping a real attack quiet at Microsoft, JPMorgan Chase and the FAA at the same time would take a secret kept by thousands of engineers, many of whom post on Reddit. (They would not keep it.)

The weak spots are worth naming: Apple's explanation is one sentence, and the FAA's root cause is still open. Neither is evidence of a cover-up.

So why does it feel like everything keeps breaking?

Because two things are true at once. The best long-term measures say failures per site are not going up. Uptime Institute, which surveys data center operators, says outage rates per site have fallen five years in a row. Parametrix, which monitors AWS, Azure and Google Cloud, counted 45 critical cloud interruptions in 2025, down from 49 in 2024, with total downtime down about 28%.

But more of the world now runs through fewer places. AWS, Microsoft and Google took 61% of the world's cloud infrastructure spending in the third quarter of 2026. Cloudflare sits in front of about 3 in 10 websites. So when one of them stumbles, thousands of businesses stumble with it, and everybody notices. Downdetector logged nearly 6 million US outage reports in the week ending November 16. (Those are complaints, not people. Still a lot of complaints.)

Think of it like traffic. There are fewer crashes per mile than there used to be, but everyone is driving on the same three highways. One fender bender and the whole region is late for work.

That's the pattern behind the whole week, from Microsoft to the FAA: one thing everything depends on, and nothing behind it when it fails.

What a small business should actually do

You can't stop Microsoft from shipping a bad update. You can decide what happens to your business when it does. Cheapest first:

  1. Write down the vendors you can't run without for a day. Phone and AI receptionist, email, website host, booking and payments, CRM, ad accounts. Next to each, the person who handles it when it breaks.
  2. Give each one a fallback. A second voice or provider for a voice agent. A second way to take a card. Call forwarding to a human, or a voicemail that promises a callback. Then test it once a quarter, because a fallback you've never tried is a rumor.
  3. Don't put the whole chain with one company. Your domain, hosting, email and DNS shouldn't all hang on one login. Keep a copy of your website, customer list and key documents somewhere separate.
  4. Keep backups you can actually restore, under different logins than the live system. AI tools and scripts with broad access have wiped live systems and their backups in one go.
  5. Subscribe to your vendors' status pages, and add simple uptime monitoring for your own site and phone line.
  6. Write a one-page outage plan. Who decides it's a vendor problem, who tells customers, what the holding message says ("our phone assistant is having trouble, please call this number"), and when to switch to the fallback.
  7. Patch calmly, not in a panic. Back up first, fix the most exposed systems first, keep a way to roll back, and don't push a change right before your busiest hour. Keep the basics against real attacks too: multi-factor authentication, prompt security updates, and staff who know a scam call when they hear one.

Want a second set of eyes on it?

If you've never written down which vendors your business can't live without, the single point of failure is usually obvious the minute you do. Bring your list (or bring nothing, and we'll build it with you) to a free 15-minute call. No pitch. You keep the plan either way.

Facts as of November 18, 2026. Microsoft's final incident report and the FAA's root cause are still pending. If either one changes the picture, we'll update this post.

Where these facts come from

Liked What You Read? Let’s Talk Strategy.

Take our free marketing assessment and find out exactly where your business can grow. No sales pitch, just real insights.