Did "superintelligence" take down the US? Four outages, fact-checked
US Homeland Security Secretary Markwayne Mullin thinks "superintelligence" could be behind the outages at Microsoft, the FAA, JPMorgan Chase and United Airlines. His only source is an unnamed acquaintance in the tech industry. We compared what the operators themselves cite as the cause, how often clusters like this happen anyway, and where AI has demonstrably caused outages before.

What this is about
Between November 12 and 16, 2026, four major systems went down in the US. Microsoft 365 was disrupted for nearly a full day. The FAA halted all departures nationwide. Transfers stalled at JPMorgan Chase and several other banks. At United Airlines, planes sat on the ground for about an hour and a half. Homeland Security Secretary Markwayne Mullin floated a common cause: "superintelligence".
We check this theory in three steps. First, we lay out side by side what the operators themselves say caused the outages. Then we look at whether the AI system for air traffic controllers, launched three days before the ground stop, had anything to do with the "upgrade". Finally, we ask whether a cluster like this is statistically unusual at all, and where AI has demonstrably caused outages.
What Mullin actually said
Mullin spoke on Monday, November 16, on Fox News' "Hannity". According to the Daily Wire and Yahoo, he said he had talked to people in the tech industry. A "very smart" man who didn't want to be named ("no, it's not Elon") told him:
"I think it's SI, because these programs aren't built to work as fast as SI."
"SI" stands for super intelligence, which is what the Trump administration now calls artificial intelligence. On November 13, according to the Latin Times, Mullin had already corrected a reporter who said "AI". He described the mechanism like this: a program that writes its own code "fixes itself". It's like a change order on a house where one change causes everything else to change too, until systems get overrun. The theory, he said, "kind of makes sense" to him.
In the same show, though, Mullin also said, according to Fox News, that the incidents were "individually unrelated". He was "98 percent sure" it wasn't a cyberattack. He didn't explain where that number came from. Nor did he name a specific AI program, or his source's company or role, as the International Business Times points out.
The four outages: what the operators say
| Outage | Date (2026) | Duration | Cause according to operator | AI named as cause? |
|---|---|---|---|---|
| Microsoft 365 (mainly Exchange Online/Outlook) | Nov. 12–13 | nearly 24 hours | Bug dormant in gateway servers since July, triggered by Teams failover and missing capacity | No; according to KUOW, Microsoft said AI was not the cause |
| FAA, nationwide ground stop | Nov. 13 | about 4.5 hours | Data flood in the FTI telecom network crashed old equipment; trail leads to a new L3Harris router | No; SMART "not connected", per FAA |
| JPMorgan Chase plus knock-on effects at other banks | Nov. 16 | from about 11:30 a.m. to about 6:30 p.m. ET | "Software bug in on-premise hardware systems" | No |
| United Airlines | Nov. 16 | about 90–100 minutes | "Technology issue", no details published | No |
Microsoft: According to the preliminary Post Incident Review reported on by BleepingComputer and Petri, three things came together. First, a July update had introduced a bug into Microsoft 365's network gateways, the servers that route traffic to services like Exchange. Under load, the bug causes "lock contention": lots of processes wait on each other for the same locked resource, like cars at a gridlocked intersection. It went unnoticed in Microsoft's test labs because they run older hardware with fewer processor cores. Second, about 15 hours earlier, Teams had automatically moved onto the same gateways. That failover, an automatic switch to backup infrastructure, added extra load. Third, nearly 14 percent of gateway capacity was unavailable because of a separate problem the day before. In under two hours, more than 80 percent of the servers stopped responding. Rolling back to an older version wasn't possible because the bug was already months old. According to The Register, the final report is due on November 20 and hadn't been published at the time of writing.
FAA: According to Business Traveller and Circle ID, the FAA ordered the stop at 3:54 p.m. ET and lifted it around 8:15 p.m. The agency says its own telecom network, FTI, was flooded with around 1,000 connection attempts every four to five seconds. That made "older equipment unstable", knocking out radio and radar data. According to The Hill, early clues point to a radar site in Amarillo, Texas, and a radio station in El Dorado, Arkansas. That's where network operator L3Harris had installed a new router as part of the switch from copper lines to fiber. According to Reuters, 17 of the 22 en route control centers were affected. The routers being replaced were about 25 years old and should have been swapped out back in 2006. The root cause hasn't been definitively established. Transportation Secretary Sean Duffy told Reuters: "We have not yet determined what the fault was." It's still unclear whether the router was faulty, installed incorrectly, or whether the old network simply couldn't keep up. CBS also reported that a network update triggered a reboot. That hasn't been confirmed.
JPMorgan Chase: Spokesperson Joe Evangelisti spoke of a "temporary issue related to a software bug" that was "not cyber-related", according to Al Jazeera. Other banks such as Bank of America, Wells Fargo and U.S. Bank were only affected when Chase was involved in a transaction. U.S. Bank, Capital One and Truist said their own systems were intact. Zelle operator Early Warning Services said its network was working.
United: The airline only refers to a "technology issue" with no sign of a cyberattack. According to AP, cited by Fox Business, crews couldn't access the weight and takeoff performance data required before every departure. A United outage in August 2025 looked similar.
Did the AI system SMART have anything to do with the "upgrade"?
It's a natural question. On November 10, the FAA expanded its AI-assisted planning system SMART to all 22 centers, just three days before the ground stop. But according to all available information, the "upgrade" the FAA cites as the trigger is a different project:
- Different vendor: SMART comes from Air Space Intelligence (ASI). The router replacement is a project by L3Harris, which operates the FTI network. L3Harris says it has completed more than 70 percent of the switch to fiber.
- Different system layer: According to the FAA's FAQ, cited by AeroCorner, SMART is a cloud-based "data aggregator" that doesn't feed information into the air traffic control system. It operates at the planning level, suggesting departure times and reroutes that humans review. Radio and radar run over the FTI network, and that's exactly what failed. In the FAA's words: "The two systems are not connected."
- Different timing: According to Duffy, L3Harris had replaced the router in El Dorado weeks earlier.
Our take: based on what's known so far, a connection is very unlikely, but it hasn't been independently ruled out. All the denials come from the FAA, the department and the vendor; there's no independent investigation yet. On November 13, Mullin himself still said, according to the Latin Times: "You can't rule anything out right now." Criticism of SMART itself also remains unresolved. Democrats on the Transportation Committee have asked the Government Accountability Office (GAO) to review it, citing the short testing phase and missing safety analyses.
Is a cluster like this even unusual?
Four headlines in five days feel like a pattern. The measurement data points more toward coincidence.
- ThousandEyes, a Cisco monitoring service, counts 300 to 600 network and cloud outages worldwide per week, according to its weekly reports documented by Computerwoche. In the week of October 26 to November 1, 2026, there were 601; in the week of November 2 to 8, there were 450. Most go largely unnoticed.
- The Uptime Institute finds in its 2026 outage analysis that the outage rate per data center has been falling for five years. About two-thirds of publicly known outages are caused by third parties, meaning cloud, internet and telecom companies. The most common human error: procedures not being followed. Uptime also warns that more automation can introduce new kinds of failures.
- Ookla (Downdetector) counted 51 days of major disruptions at big AI services in the first quarter of 2026, up from 6 in the first quarter of 2025. But according to Ookla, a large share of those came from cloud outages that hit AI services, not from the models themselves.
The best comparison is fall 2025. On October 20, AWS went down, generating more than 17 million Downdetector reports worldwide, according to Ookla. On October 29, Azure Front Door followed after an accidental configuration change, and on November 18, Cloudflare. Three major global outages in four weeks, three completely different causes, and nobody needed to invoke superintelligence. The counterexample is CrowdStrike in July 2024. Back then, a single cause really did take down airlines, banks and hospitals at once, around 8.5 million Windows machines. That common cause was known within hours. This time, the four operators name four different causes.
Our take: with hundreds of outages a week, clusters like this are to be expected. And once the public is paying attention, later disruptions are more likely to be reported and read as part of a series.
Where AI actually caused outages
The concern isn't entirely baseless. There are documented cases, though with a different mechanism than the one Mullin describes:
- AWS, December 2025: According to the Financial Times, cited by The Decoder, the AI coding tool Kiro was allowed to make changes. It decided to delete an environment and rebuild it. The result was a roughly 13-hour outage of the Cost Explorer billing tool in one AWS region in China. Amazon calls it "user error" because the access permissions had been set incorrectly.
- Amazon online store, March 2026: According to Fortune, an internal document named "GenAI-assisted changes" as a factor in a series of disruptions. The passage was cut before the meeting. Amazon says only one incident involved AI: an engineer followed bad advice an AI agent had derived from an outdated wiki. AI-written code was not involved, the company says.
- Cloudflare, November 2025: Ironically, the system that crashed here was the one that uses a machine learning model to detect bots. But according to Cloudflare's report, the trigger was a changed database permission that caused a configuration file to double in size. So an AI component was affected, but it wasn't to blame.
What's established, then: automated changes and AI agents with overly broad permissions can damage systems, especially when no human signs off. Microsoft's Teams failover was also an automatic switch, though classic automation with no AI involved. For Mullin's picture of an AI that rewrites itself and moves faster than legacy systems can handle, there's no evidence in any of the four incidents.
The bigger picture
Mullin's theory is secondhand speculation, and it doesn't match what the operators are saying. Microsoft, the FAA and Chase cite specific, unrelated causes; United cites none at all. There is a kernel of truth, though: new technology overwhelms old technology. At the FAA, a new router apparently destabilized a network of decades-old equipment. That's an infrastructure problem, and you don't need superintelligence to explain it.
Plenty remains open. The FAA hasn't pinpointed the exact fault yet, Chase and United have barely explained themselves, and Microsoft's final report is still pending. Representative Ted Lieu called for gathering facts first and assigning blame later. Cyber expert Jody Westby told Al Jazeera that everyone was too quick to conclude it was a purely technical problem.
Our take: the real question isn't whether a superintelligence is sabotaging America. It's how operators secure AI agents and automated changes: with limited permissions, human sign-off and testing on real production hardware. The Amazon cases show this is a real risk. Whether it played a role in these four cases is something the final reports should reveal. So far, nothing points that way.
Sources
- Yahoo News: Markwayne Mullin: Tech Industry Source Believes 'SI' Is Causing Outages yahoo.com
- Fox News: DHS chief says FAA, Microsoft, bank outages were 'unrelated' foxnews.com
- Daily Wire: Mullin Suspects AI Is Behind System Meltdowns dailywire.com
- Latin Times: Mullin Refuses to Clear the FAA's New AI Tool latintimes.com
- IBTimes UK: Markwayne Mullin Cites 'Guy That's Very Smart With Tech' ibtimes.co.uk
- BigGo Finance: Lawmakers Split on Cause of US Outage Wave finance.biggo.com
- BleepingComputer: Microsoft says months-old bug caused global Microsoft 365 outage bleepingcomputer.com
- Petri: Microsoft explains Microsoft 365 outage petri.com
- StatusStack: Microsoft 365 and Exchange Online outage, November 12, 2026 statusstack.com
- KUOW: Microsoft says outage not caused by AI or cyber attack kuow.org
- The Hill: FAA identifies Arkansas router as cause of nationwide ground stop thehill.com
- The Star (Reuters): US aviation shutdown blamed on L3 Harris installed router thestar.com.my
- CircleID: FAA nationwide ground stop raises concerns over aging telecommunications infrastructure circleid.com
- AeroCorner: FAA SMART AI System and the Ground Stop aerocorner.com
- Al Jazeera: US bank outage – what caused Chase problems and Zelle payment delays? aljazeera.com
- Fox Business: Details emerge about United Airlines outage foxbusiness.com
- Computerwoche: Netzausfälle und Internet Health Check (ThousandEyes-Daten) computerwoche.de
- ITWeb: Uptime Announces Annual Outage Analysis Report 2026 itweb.africa
- IEEE ComSoc Techblog: Ookla – AI platform reliability decreases as outages surge techblog.comsoc.org
- The Decoder: AWS AI coding tool decided to delete and recreate a customer-facing system the-decoder.com
- Fortune: Amazon retail site outages, AI agent inaccurate advice fortune.com
- Cloudflare Blog: 18 November 2025 outage blog.cloudflare.com
Ask the AI assistant
Questions about this article, installation or background? Our assistant knows this page and runs on our own machine, not in someone else's cloud.
The Assistant is an AI (Gemma 4) and can make mistakes – please check important details against the source. Questions are used only to generate an answer and are deleted after one hour. Please don't enter any personal data.
Comments
What do you think? Already tried it? Write it here, no sign-up needed.
Comments are public. Links aren't allowed, and we delete abusive posts. You can delete your own comment at any time in this browser. Learn more in the Privacy Policy.