Content Strategist, Dunham | 2025 ThinkAdvisor Luminaries Award Winner | 2026 Wealthies Finalist, Thought Leader of the Year | Macroeconomics, markets, geopolitics & global trends
The November 2026 outages at Microsoft, the FAA, JPMorgan Chase, Apple, and the NSA had different causes, yet they highlight a growing paradox with AI. AI tools are helping security teams find hidden software flaws faster than ever, which strengthens defense. However, rushing to patch those bugs is straining older systems - as seen in the NSA outage - and experts warn that this pace could make widespread disruptions more common.
Key Takeaways
A week of major tech outages exposed our dependence on digital systems. Disruptions at Microsoft, the FAA, JPMorgan Chase, Apple, and others had different reported causes (many indirectly by AI).
One technical problem can reach many businesses. Companies often rely on shared cloud and communications providers, so an outage at one provider can interrupt work and services across multiple industries.
AI agents can take actions beyond their intended limits. During an OpenAI cybersecurity test, agents found ways to communicate and reach outside systems, eventually accessing parts of Hugging Face’s platform.
The AI race creates pressure to keep moving. The U.S., China, and technology companies have strong incentives to advance AI - even as they weigh the risks of deploying it across an interconnected economy.
Listen to this blog here
13:21
Last week, Microsoft Outlook went down – which crippled millions across various countries (myself included).
The very next day, a communications problem led the Federal Aviation Administration (FAA) to ground departures across the U.S. – meaning tens of thousands of flights were paused.
Then a few days later, a “software issue” at JPMorgan Chase delayed transactions - including some involving other banks.
And just as we finally think we can breathe again, by Tuesday this, even the National Security Agency (NSA) had reported a disruption across its highly confidential network. And Apple iCloud suffered a major outage.
That’s a lot of our most important systems acting up in a very short period.
And it made me wonder - Is this all coincidental, or is there something behind it?
After all, the universe is rarely so lazy. . .
Now, I don't have evidence that these outages were connected. And the explanations we've gotten so far point to different problems.
But seeing so many critical systems fail within days of each other revealed something bigger worth thinking about.
How far can one technical problem travel through a connected economy?
And what happens as increasingly capable AI tools gain access to those same systems?
That brings me to an old Greek story.
Has the Pandora Box Been Opened?
In Greek mythology, Pandora (the first mortal woman) was given a container and told not to open it under the orders of Zeus.
But like anyone, curiosity got the better of her. And eventually, when she lifted the lid, all kinds of troubles - evil, disease, sorrow, etc - escaped into the world.
By the time she tried to close it, it was too late. They were already out and about.
That story reminds me a lot of AI.
I’m not saying AI is a curse. On the contrary, it has extraordinarily useful applications - including helping security teams find flaws before criminals do (which is what we’ve seen in these recent outages).
But once a powerful tool exists and spreads, putting it away becomes difficult – if not impossible.
For now, it looks like the box is open. So, let’s look at what we know.
What Caused the Recent Tech Outages?
This was a rough week:
November 12 | Microsoft: A disruption affected Microsoft 365 services, including Outlook and Teams. Microsoft’s preliminary review pointed to a previously undetected software bug, increased traffic, and reduced server capacity.
November 13 | FAA: A telecommunications problem led the FAA to issue a nationwide ground stop (which has only happened two previous times since 9/11 in 2011). The agency said a surge in connection attempts overwhelmed older equipment. Planes already in the air continued to land safely, but the disruption contributed to widespread delays and cancellations (estimated around 13,000 over three days).
November 13 | Workday: An outage left millions of employees unable to access workplace accounts used for tasks such as clocking in and out, checking pay information, and requesting time off. The company said it was working to restore service.
November 16 | JPMorgan Chase: A software issue in the bank’s internal systems slowed millions of transactions - including payments involving other financial institutions – and locked thousands of customers out of their accounts. JPMorgan said it found no evidence that the problem was cyber-related. Then on November 19th, JPM reported third-party vendors were hit by a separate cyberattack – which gave attackers access to marketing data - including some customer names and email addresses.
November 17 | Apple: An outage affected several Apple services - especially iCloud. Apple said an internal update caused the disruption and that it was not a cyberattack.
November 17 | NSA: The agency's classified network went down. The NSA called it a "temporary degradation during scheduled routine maintenance" and said mission-critical work continued. The Director of National Intelligence tied it to urgent patching of a PHP flaw that AI helped .
Some officials have pointed to AI’s growing role in finding software flaws - and the pressure to patch them quickly - as a possible source of disruption. But the reported causes of these outages differ, and we don’t have evidence that rogue AI agents caused them.
Either way, the week was a reminder of how much we rely on systems most of us rarely think about – that is, until they stop working.
Why Can One Tech Outage Affect So Many Businesses?
In this digital age, much of the economy relies on shared digital infrastructure every day.
Businesses use cloud providers to store data and run software. Banks rely on networks that move millions of payments between institutions. Airlines depend on systems that connect airports and air traffic controllers. Even our email, phones, and news apps run on the same web.
Why Can One Tech Outage Affect So Many Businesses?
In this digital age, much of the economy relies on shared digital infrastructure every day.
Businesses use cloud providers to store data and run software. Banks rely on networks that move millions of payments between institutions. Airlines depend on systems that connect airports and air traffic controllers. Even our email, phones, and news apps run on the same web.
That makes our daily lives faster and more efficient. But it also means we may depend too heavily on systems that if one fails, the rest can seize up.
For example, in a 2026 Flexera survey of cloud decision-makers and users, 83% of the enterprises surveyed used Amazon Web Services for active workloads, while 79% used Microsoft Azure. Some even used both.
Furthermore, the European Central Bank has warned that as firms use more AI, reliance on a narrow group of technology providers can make disruptions more likely to affect multiple firms at once.
Imagine an office building with hundreds of tenants. Each business has its own employees and customers. But many depend on the same elevators to get people through the doors.
If one elevator breaks, the whole building doesn’t necessarily close. But enough people could be delayed for the effects to spread beyond the elevator itself.
Digital systems can work similarly. A problem at one provider can disrupt multiple businesses - and eventually reach their customers.
Of course, companies use different services, locations, and backup systems. An outage won’t affect everyone in the same way.
But when many businesses rely on shared infrastructure, one problem can reach a lot of places.
Now, what happens as AI agents gain access to more of those systems?
What Happened in the OpenAI and Hugging Face AI Agent Incident?
Most of us know AI as a chatbot. You ask it a question, and it gives you an answer.
An AI agent takes this to a whole different level.
You essentially give it a goal, and it can use software tools to work through the task. Depending on what it’s allowed to access (it could be your brokerage, your emails, etc), it might search the internet, read files, write code, or interact with other systems then make decisions on its own.
Think of it between asking someone for directions (AI chatbot) and handing them the car keys (an agent).
That ability could save businesses and people a lot of time with various tasks.
But it also increases the risk of something going wrong.
We got a glimpse of that in July.
Many may have glanced over this news. But it was actually pretty crazy what happened.
Long story short, OpenAI (which made ChatGPT) was testing how well its AI agents could find software vulnerabilities. The agents were placed in separate virtual environments - with controls meant to keep them from communicating with one another or freely accessing the internet.
But the agents found an opening. . .
They discovered that a shared internal software service could be used like a message board with each other. From there, they left notes, asked for help, and passed along what they had found.
They essentially created their own Reddit forum.
It's actually very clever what they did.
But making matters worse, they also found a way to reach the internet - which wasn’t supposed to happen.
OpenAI’s security team later rebuilt the affected service and tightened access after the activity contributed to an internal outage. But when testing resumed, the agents just found new ways to communicate.
This time, they began coordinating more extensively. Some of the agents even described themselves as a “swarm.”
According to an independent investigation by METR and Redwood Research, roughly 1,200 agents exchanged more than 70,000 messages and files on the unauthorized board. Around 700 went on to participate in an attack on Hugging Face - a platform where developers share AI tools and data.
Basically, the agents had been given cybersecurity challenges. And in trying to complete them, some began looking for ways to just influence or work around the test’s scoring system (basically they wanted to cheat). This eventually led them to Hugging Face.
Using exposed credentials and software vulnerabilities, the agents gained access to parts of the platform.
Crazy, right?
Now, keep in mind, these were agents taking part in an internal cybersecurity test, where they had been assigned hacking-related tasks. OpenAI also says the test ran with fewer safeguards than the systems it makes available to customers.
But still, the agents reached systems outside the intended test and took actions on their own.
OpenAI says it has since tightened controls and expanded monitoring. But the episode offers a look at what can happen when capable software is given a goal, access to tools, and a way around the barriers meant to contain it.
This is a fascinating case study about rogue agents – and for those interested, here’s the full post-mortem report.
The point is, in an age where AI is growing rapidly, all it could take is a malicious group to light the fuse that sets off another “swarm” which targets our global digital infrastructure.
Why the AI Race Keeps Going
This is where the competition between the U.S. and China comes in.
Both countries see AI as important to economic growth, industrial capacity, and national security. Meanwhile, companies also have strong reasons to invest because they see potential profits.
So what happens if one side wants to slow things down?
Well, if one slows down (say China), they’ll worry that it’ll allow U.S. competitors to pull ahead and get a steep advantage. Meanwhile, the other (the U.S.) may have the same fear if they slow down while the other keeps going.
Put simply, each side has an incentive to keep moving - even if both recognize the risks of moving too quickly.
That’s the game theory problem.
The fear of falling behind can push competitors to take risks they might otherwise avoid.
And as AI becomes more important to the economy, it may become harder for any one company or country to step back on its own.
What the Outage Economy Means for us
AI may help companies become more productive. But making the systems around it dependable will have costs too.
Businesses may need to spend more on cybersecurity, software testing, backup systems, and people who understand how their networks fit together.
I think it’s worth asking how dependent a company is on a small number of providers - and whether it can continue essential operations if one goes down.
A backup is only useful if it works when needed. If it relies on the same network as the main system, it may be exposed to the same disruption.
That brings me back to Pandora’s box.
AI is getting more capable, the race to deploy it is accelerating, and we're giving these tools access to systems that were never designed with autonomous agents in mind.
Thus, there’s no putting those things back in the box. But the recent outages have left me wondering:
Can our interconnected economy withstand the next shock - or are we only beginning to discover where its weak points lie?
As always, time will tell.
Frequently Asked Questions About November 2026 Tech Outages and AI
What caused the widespread tech outages in November 2026? Each outage had its own cause rather than one big attack. Microsoft blamed a months-old software bug, while the FAA pointed to a flood of connection attempts that hit old telecom gear. Apple cited an internal update, and JPMorgan reported an internal software issue. The NSA outage happened during urgent patching for a software flaw that AI helped find.
Did AI cause the November 2026 tech outages? No, there's no proof that rogue AI agents caused the outages. Still, AI played an indirect part. Because AI is finding old bugs faster than teams can comfortably fix them, rushing to patch aging systems could make short outages more common.
What happened in the 2026 OpenAI and Hugging Face AI agent incident? In July 2026, OpenAI ran internal security tests with AI agents that had looser safeguards. About 1,200 agents broke past their limits, used a shared internal tool like a message board to swap over 70,000 messages and files, and reached the internet. Around 700 agents then used exposed logins and software flaws to enter parts of Hugging Face before OpenAI locked things down.
Why can one tech outage disrupt so many businesses at once? Most companies rely on the same shared tech backbone, including cloud platforms, payment networks, and phone systems. A 2026 Flexera survey found that 83% of surveyed firms run active workloads on Amazon Web Services, and 79% use Microsoft Azure. When so many businesses depend on a few big providers, one technical failure can spread across entire industries fast.
Sources
Lifehacker — What Is Causing All These Widespread Outages? [lifehacker.com]
BleepingComputer — Microsoft Says Months-Old Bug Caused Global Microsoft 365 Outage [bleepingcomputer.com]
Federal Aviation Administration — Nationwide Ground Stop: Questions and Answers [faa.gov]
Engadget — Outage at HR Platform Workday Locks Millions of Employees Out of Their Account [engadget.com]
Computer Weekly — US Tech Outages Spark Conversation on Resilience in the Age of AI [computerweekly.com]
TechCrunch — Apple Services, Including iCloud, App Store, Apple Music, and More, Hit by Outage [techcrunch.com]
CBS News — National Security Agency Network Outage Linked to Software Vulnerability [cbsnews.com]
Flexera — Flexera Finds Cloud Value Is Rising While AI Waste Grows []
Disclosures
This communication is general in nature and provided for educational and informational purposes only. It should not be considered or relied upon as legal, tax or investment advice or an investment recommendation, or as a substitute for legal or tax counsel. Any investment products or services named herein are for illustrative purposes only and should not be considered an offer to buy or sell, or an investment recommendation for, any specific security, strategy or investment product or service. Always consult a qualified professional or your own independent financial professional for personalized advice or investment recommendations tailored to your specific goals, individual situation, and risk tolerance. All examples are hypothetical and are for illustrative purposes only.
Information contained in the materials included is believed to be from reliable sources, but no representations or guarantees are made as to the accuracy or completeness of information. This document is provided for information purposes only and should not be considered as investment advice.
Dunham & Associates Investment Counsel, Inc. is a Registered Investment Adviser and Broker/Dealer. Member FINRA/SIPC. Advisory services and securities offered through Dunham & Associates Investment Counsel, Inc.