Why We May Be Under Attack and Not Yet Know It
Chase blames a software bug it hasn't named. United says — three times — there's no indication of a cybersecurity issue, and hasn't named the system that failed. Wells Fargo has offered no cause at all. Here's how penetration testing and hostile probing of live systems actually work, what Monday's confirmed record shows, and the questions the official explanations leave open.
Edited and fact-checked by Stephen Sprague
Every owner graded its own homework on Monday, in the first hours, before any forensic work could possibly be done. The grades may hold up. They haven't been checked yet.
Monday produced a cluster that anyone who runs systems would circle on a calendar. JPMorgan Chase, the country's largest bank, spent most of the business day with its transaction processing slowed. Wells Fargo acknowledged technical issues on its own website and offered no cause at all. By evening, United Airlines couldn't get dispatch data to its departing flights for about an hour. Three systems, no shared owner, two sectors, one business day — and three official explanations that don't yet explain much.
The companies have answers, and this page prints them in full below. But a first-hours statement from the owner of a failed system is a claim, not a finding. The owner is grading its own homework, on a deadline, with its share price and its regulators watching. That doesn't make the claim false. It makes it the start of the questions, not the end of them. To ask them properly, it helps to know what the thing people are quietly wondering about — a penetration test, run by someone who never asked permission — looks like from the outside.
A penetration test is an authorized rehearsal for an attack: an organization hires specialists to break into its own systems the way an intruder would, so the holes get found by friends first. Hostile probing runs the same craft without the invitation. Both start the same way — not with a battering ram, but with a notebook.
The first phase is reconnaissance, and most of it is quiet. The tester maps the target: which systems talk to which, where a single failure would travel. Dependencies are the real prize. A modern bank or airline isn't one system; it's a web of them, and the map of that web — who slows down when a counterparty slows down, which application feeds the departure board — is worth more than any single password. You don't learn that map from a brochure. You learn it by touching things and watching what moves.
Then comes the probing itself: small, deniable touches, each one a question. How fast does the owner notice? How fast do they recover? What do they say in public, and how much of it holds up? The response is the intelligence product. A probe that knocks something over and watches the recovery has learned the owner's detection time, the strength of their backups, and whether a failure stays contained or ripples outward to everyone connected to the target. The prober doesn't have to win. They have to learn.
The best probing doesn't look like an attack at all. It uses the system's own tools, its own accounts, its own ordinary traffic — a technique investigators call living off the land — so the activity blends into the noise every large system makes on a normal day. Which is why an owner's first-hours answer is narrower than it sounds. "No evidence of a cyberattack" means no evidence has surfaced yet, in the places they've looked so far. It's a statement about visibility. It is not a forensic conclusion, and it can't be, hours in.
This isn't theory. In February 2024, CISA, the NSA and the FBI, joined by allied agencies, published a joint advisory on Volt Typhoon, a group the agencies attribute to China's government. They confirmed the group had gotten inside the IT environments of multiple U.S. critical infrastructure organizations — mainly in communications, energy, transportation and water — and had kept access in some victim environments for at least five years. The method was living off the land: legitimate tools, valid accounts, activity that read as routine administration. The agencies assessed the group wasn't primarily stealing secrets; it was pre-positioning, positioned to disrupt those systems in a crisis. Microsoft first described the campaign publicly in May 2023. For years before that, it looked like nothing at all.
That precedent is the lens for Monday. The last confirmed case of hostile actors inside American infrastructure went undiscovered for years precisely because it was built to be indistinguishable from ordinary system noise. That history doesn't make any given outage an attack. It does mean "we see no evidence" — offered in hour one, by the owner, about its own systems — has a documented failure mode, and a skeptical reader is right to price it accordingly.
Now the day itself. Chase's customers began reporting trouble at about 9:53 a.m. ET. The bank's public statement came at 3:57 p.m.: transaction processing had slowed because of "a software bug in on-premise hardware systems"; there was "no evidence this is cyber-related"; and it was "not a systemic issue" — other banks, Chase said, may have seen slowness "only to the extent that they were a party to a transaction with us." Shortly after 7 p.m., transactions were processing normally again. Print the explanation. Then look at what it asks you to accept. The bug is unnamed. The hardware is unnamed. No account of the failure's path inside the bank has been published. And the one detail Chase did offer — that other banks slowed exactly where they touched Chase — is a dependency map drawn in public: a failure inside one institution, propagating along its connections in a pattern anyone watching could chart. That's the innocent reading, and it may be the whole story. It's also, not coincidentally, the thing a probe exists to learn.
Wells Fargo is the loudest silence of the day. Reports of trouble surfaced around 12:46 p.m. The bank acknowledged "technical issues" on its own website — and has said nothing about a cause at all. It hasn't blamed a bug. It hasn't ruled anything out. It hasn't said whether its problems connect to Chase's. There may be a dull explanation. It hasn't been provided, and until it is, Wells Fargo's column in Monday's ledger is blank — which is itself a fact worth printing.
United's window was the evening. Pilots were reporting to air traffic control by about 4:02 p.m. that they couldn't receive final weight-and-balance and takeoff data — the numbers a crew needs before it can legally depart. The airline confirmed a widespread system outage just before 5 p.m., said flights in the air and United Express flights weren't affected, and announced the issue resolved at about 6:01 p.m. The day's final tally, from Reuters' FlightAware count, was 645 delayed flights — about a fifth of United's schedule — and 25 cancellations. United's statement on cause, repeated three times across the evening: "There is no indication that this was a cybersecurity-related issue." Our full report on the outage is linked below. Note what the statement doesn't contain: the name of the system that failed, why a technology outage reached the dispatch data a departure depends on, or any account of what the airline's own logs show. "No indication" is doing all the work in that sentence — and it was doing it within hours of the failure.
One event from Monday doesn't belong in this ledger, and it matters to say so. Amtrak's service around Washington was halted for hours because a train struck a trespassing individual near New Carrollton and local officials held the tracks. That's a physical tragedy, not a systems event, and it earns no place in a pattern about software. Connecting dots honestly means setting aside the dots that don't connect.
And Monday isn't the whole of the question. The pattern that raises it spans six weeks, and the confirmed record — built from the owners' own status pages and statements, not the rumor mill — runs like this. On Thursday, November 12, Microsoft 365, the email and office backbone for a large share of American business, went down globally in an outage that ran into Friday; Microsoft attributes it to servers becoming unexpectedly unavailable under shifting load, and says it found no evidence of malicious activity. That Friday, November 13, the FAA stopped air traffic nationwide for over four hours — only the second nationwide ground stop since 9/11 — after a new router installed by contractor L3Harris during a copper-to-fiber conversion flooded the network with connection attempts; the agency's formal root-cause determination is still pending. On Sunday, November 15, Amazon Web Services logged an incident in its US-EAST-1 region from 6:05 to 9:21 a.m. ET — sign-in failures and connectivity problems across 11 services — and AWS has published no cause at all. Then Monday: a bank, a second bank, an airline.
Underneath the headline events sits a quieter drumbeat, and it belongs to the Federal Reserve. The Fed's own status page records the Fedwire Securities and National Settlement services running degraded for roughly 30 hours across the weekend of November 7–8 — and shorter degradations on November 1, 2, 6, 9 and again on the evening of November 12. Every one is officially acknowledged. Not one carries a published cause. Two more majors belong in the same pile: on October 15, a network disruption at Nasdaq's Carteret, New Jersey facility knocked out U.S. equities data feeds across all three of its exchanges for hours, and Nasdaq has not disclosed a cause; and on October 20, Google Cloud lost power completely at the London facility hosting one of its zones — an underground cable fault, on Google's own detailed account, with both emergency-generator controllers failing to start — while UK financial-sector firms were among the customers whose systems went down.
The counterweight held for all six weeks, and it deserves its full weight. The systems that clear everyone else's business never failed: Visa and Mastercard declared no outage of their networks at any point in the window, Zelle ran, FedNow ran — including all day Monday — and Gmail posted no incident in six weeks. No other major airline had a systems outage; the carriers' only mass disruption came as victims of the FAA's stop. Whatever was failing in those six weeks, it was never the connective tissue itself.
Around the majors sits the ordinary weather of very large systems — a regional Azure failure in Britain, a European payments hiccup, minor status-page entries at Stripe, Apple and Cloudflare, and three 911 failures in California and Oregon that share one mundane mechanism: stolen copper and cut fiber. One pattern in the window is confirmed and openly criminal — and separate. Four ransomware intrusions hit American hospitals in four weeks, and a breach at Finzly, a payments-technology vendor, exposed files that included the configuration of the firewall guarding its bank clients' link to the Federal Reserve. No outage followed; the Fed says its payment services were not impacted; and no official source connects any of it to the failures above. It goes in a different ledger — but it establishes that in the same six weeks, hostile actors really were inside American systems. Just not, on any evidence produced so far, inside the ones that failed.
So the six weeks resolve into two piles. In one: explanations, printed on this page as given — Chase's unnamed bug, United's 'no indication,' Microsoft's server account, the FAA's contractor account, Google's cable fault. In the other: the residue nobody has explained at all. AWS has published no cause. The Federal Reserve has published no cause for any of its six degradations. Nasdaq's cause is undisclosed. Wells Fargo has given no cause. And the FAA's formal determination — the one finding in the window that would come from investigators rather than from the owner — is still pending. The window's most consequential open items rest on either an interested party's word or on silence. That is not an accusation. It is an inventory — and inventories are supposed to balance.
Some of those entries arrived with an official explanation attached, and each explanation may be true. But notice the shape of the six weeks: the failures keep landing in the systems other systems stand on — email, air traffic control, cloud infrastructure, payments, dispatch. Infrastructure doesn't have to be attacked to be mapped. It only has to be watched while it falls down on its own — if that's what it's doing.
The pattern also has to survive its strongest counter-evidence, and here it is. The systems a coordinated probe of American finance would almost have to touch stayed officially normal all day. Visa said its systems were operating normally. Mastercard said its network was operating normally. Zelle's operator said, "Zelle is not down. Zelle is operating as it should." Bank of America and U.S. Bank said their own systems weren't directly impacted — only their Chase-facing transactions slowed, which matches Chase's account. Gmail posted no incident on Google's status dashboard, and Microsoft 365's major outage was last week's, not Monday's. If Monday was a test of the connective tissue, the tissue reports it was never touched. That's a real fact, and it cuts against the pattern — which is exactly why it belongs on this page.
So run the numbers, honestly, in the direction they cut hardest against the pattern. Count the American institutions big enough to make national news when a system fails — the major banks, the card networks, the big airlines, the cloud providers, the email platforms. There are roughly forty. If each produces one publicly noticed failure a year — a conservative rate, given what status pages show — a normal year already contains about forty of them, close to one a week. Six weeks, by chance alone, expects about six. And chance doesn't deal events evenly: random failures clump, the way random raindrops clump, and even spacing would be the anomaly. Failures also echo. Chase's single bug manufactured apparent outages at every bank that touched it — one event, wearing five costumes, all over the day's outage trackers.
Now the other cut, because the math has a limit and the limit matters. Frequency prices how often, never what. It can't weigh which functions failed: transaction processing at the country's largest bank, and dispatch — the data a departure legally depends on — at one of its biggest carriers. Those are the two systems a prober maps first, because everything else hangs off them. And frequency can't price the prior: a confirmed, documented campaign, named by three U.S. agencies, that sat inside sectors like these for five years and looked like nothing. Base rates describe the average week. They don't testify about this one.
History rhymes in both directions, and both rhymes are on the record. In July 2024, a Microsoft Azure outage and CrowdStrike's bad update landed inside 48 hours of each other; airlines grounded, banks and 911 lines failed around the world, and for a day it felt exactly like coordination. It was one bad file, plus noise. And then there's Volt Typhoon — the time the quiet pattern was real, and read as nothing for five years. The numbers say a cluster like Monday's can happen by chance. They do not say this one did. Anyone who tells you the math settles it, in either direction, is selling something the math doesn't contain.
There is a reason the agencies chose the word they chose. The Volt Typhoon advisory doesn't accuse the group of spying, and it doesn't accuse it of stealing. It says the actors were “seeking to pre-position themselves” on American infrastructure networks “for disruptive or destructive cyber activity” “in the event of a major crisis or conflict.” Read that the way a planner reads it. The access isn't the operation. The access is the option — bought quietly, held patiently, worth nothing until the day it's worth everything. Pre-positioning is what comes before an attack. It's the phase an attack is built out of before the attack exists, and by design it produces no event to point at: no outage, no ransom note, no headline. Just presence, buried in the ordinary untidiness of very large systems.
The one time the sequence ran to completion in public, it ran exactly like that. On December 23, 2015, remote intruders opened breakers at three Ukrainian power companies and cut electricity to roughly 225,000 customers — the first confirmed cyberattack to black out a power grid. The Department of Homeland Security's investigators concluded the strike was “synchronized and coordinated, probably following extensive reconnaissance of the victim networks.” The intruders had stolen legitimate credentials in advance, and the malware found on the utilities' systems, the alert said, most likely did its work “in the reconnaissance and preparatory phases.” The public reporting filled in the rest: months spent inside, learning how the control systems actually ran, before a single breaker moved. Months of learning. One winter afternoon of switching. The learning was the precursor; the switching was just its last step.
Now the second ledger: the threats the United States has already had, in the open, in court. In November 2024, the Justice Department unsealed a complaint against Farhad Shakeri, described by prosecutors as an asset of Iran's Islamic Revolutionary Guard Corps living in Tehran. According to the complaint, Shakeri told law enforcement in recorded interviews that the IRGC had tasked him on October 7, 2024 with providing a plan to kill President-elect Donald Trump — and that he supplied the Guard with operatives drawn from a network of criminal associates he met in American prisons. Two of them, Carlisle Rivera and Jonathan Loadholt, spent months surveilling journalist Masih Alinejad, whom they'd been promised $100,000 to kill; both have since pleaded guilty and gone to prison for it. Shakeri remains at large in Iran, and the charges against him are allegations he has never answered in an American court. In a separate case, a federal jury convicted Asif Merchant in March 2026 of murder-for-hire and attempted terrorism in another Iran-backed plot to kill American politicians. These aren't scenarios from a war game. They're case numbers.
The threat stream isn't history. Last Friday, the Secretary of the Treasury confirmed on CNBC that he has moved out of his Washington home and onto a military base. “I follow the Secret Service instructions and have moved from my home to a military base,” Scott Bessent said, calling Iran “a dangerous cornered animal” and the death threats against him retaliation for the sanctions campaign he runs. Reuters reported that a senior intelligence official had warned in a memo weeks earlier that Iranian plots against American personnel and citizens had reached an advanced stage. Iran has denied plotting attacks inside the United States. But a sitting cabinet secretary is sleeping on a military base because the Secret Service told him to. That's not a think-tank scenario. That's this month.
Iran's cyber arm has already touched the two kinds of systems that failed on Monday: finance, and the controls underneath physical infrastructure. In 2016, the Justice Department unsealed an indictment charging seven Iranians employed by two IRGC-sponsored companies with a coordinated campaign, running from late 2011 to mid-2013, against 46 companies — most of them American financial institutions, Bank of America, JPMorgan Chase and Nasdaq among them. The attacks ran on more than 176 days, knocked hundreds of thousands of customers out of their accounts and cost the banks tens of millions of dollars. One of the seven, Hamid Firoozi, was also charged with breaking into the control systems of the Bowman Avenue Dam in Rye, New York, in August and September 2013, where he could read the dam's water levels, its temperature and the status of its sluice gate. What he didn't know was that the gate had been manually disconnected for maintenance; otherwise, prosecutors say, his access would have let him operate it. The U.S. Attorney called the dam infiltration “a frightening new frontier in cybercrime.” That was a decade ago. The frontier hasn't moved back.
Then, Monday evening, CNN reported something that belongs on this page with its fences up. CNN says new U.S. intelligence indicates Iran is developing plans to target senior Trump administration officials by detonating a dirty bomb or a chemical weapon in Washington. It cites seven U.S. officials, none of them named. Those officials describe options being built — not an attack decided, and not one they call imminent. Two of them described the plotting as specific, with locations and targets, and said it had triggered contingency planning. One cited intercepted Iranian chatter claiming bomb-making materials are already stashed inside the United States; CNN says officials are divided on that claim. The Office of the Director of National Intelligence declined to comment. The White House's on-record answer, from spokesperson Lauren Bis, is part of the record and is printed here in full: the United States is “monitoring any and all potential threats, but at this time there are no credible threats to the homeland.” Iran's mission to the United Nations categorically rejected the chemical-weapons allegation as “false and fabricated”; it did not address the dirty-bomb planning. TML hasn't seen this intelligence and can't verify the officials' account of it. It's printed as what it is: a major American news organization reporting, on the same day three systems failed, that unnamed officials describe an adversary building options for something worse — while the White House says, on the record, that no credible threat exists.
Why would a state prepare this way? Because for Iran, there is no other way to reach. The conventional ledger isn't close. By SIPRI's count, the United States spent $997 billion on its military in 2024; Iran spent $7.9 billion — roughly 126 American dollars for every Iranian one. Iran has no aircraft carriers and no global network of bases. Its air force still flies jets delivered before 1979. Its longest-range missiles can strike American targets and allies in its own region — and cannot reach the American homeland. The intelligence community's 2025 Annual Threat Assessment frames Iran as a state that challenges the United States asymmetrically while avoiding direct war, and says of its cyber arm that Iran's “increasing expertise and willingness to conduct aggressive cyber operations pose a significant threat to the security of U.S. networks and data.”
So look at which Iranian instruments have actually reached the American homeland in the public record. Not a fleet. Not a missile. A botnet that knocked bank customers out of their accounts for the better part of two years. A hacker reading a dam's control screens in Westchester County. A deported ex-convict in Tehran running prison friends by phone against American targets. Every Iranian instrument that has ever touched the United States arrived through a wire or a hired hand — because those are the only instruments Iran owns that can make the trip. For a state in Tehran's position, cyber pre-positioning isn't one option on the menu. It's the only line on the menu that reaches. And if a crisis ever comes, the preparation for it won't look like mobilization. It will look like the quiet phase. Like access. Like learning. Like nothing.
So, plainly, what this page is and isn't saying. It isn't saying Monday was an attack. No official source attributes Monday's failures to Iran, to China or to anyone else, and nothing here does either. What it is saying is that three things are true at the same time, on the record. The agencies themselves confirm that one hostile power spent years pre-positioned inside American infrastructure, in the very sectors that failed this week. The courts and the Treasury Secretary confirm that another is running assassination plots against American officials and threatening a cabinet secretary into military housing — and Monday brought a report of that same power weighing weapons of mass effect against officials in the capital, a report the White House answered by saying no credible threat exists and the intelligence agencies wouldn't discuss at all. And when the load-bearing systems failed anyway, the explanations the public got were written by the owners, in the first hours, and checked by no one. Those facts don't add up to a conclusion. They add up to a posture. The posture ought to be awake.
This is dangerous, and it isn't a game. The systems that stumbled across the last six weeks are the ones everything else stands on — the movement of money, the email, the airspace, the cloud, the dispatch desk. If they're failing on their own, the country deserves the receipts that prove it, because “on their own” is doing enormous work in that sentence. If they aren't failing entirely on their own, then the first-hours denials are exactly what a real precursor would produce while it watched them being accepted. Either way, the questions have to get tougher — of the owners, who owe named systems, named causes and timelines that can be checked, and of the agencies, the only institutions that can see across owners, who owe the public an attribution or an all-clear instead of silence. A country that can't tell the difference between a bad week and a rehearsal has already handed the rehearsal its cover.
So the questions, pressed plainly. Why did three systems with no shared vendor, no shared network and no shared owner fail inside one business day, in two of the very sectors — transportation among them, and the financial system beside it — that the Volt Typhoon advisory names as pre-positioned targets? Why does the public record, a day later, still consist of three acts of self-assessment, two silent about the actual failed component and one empty altogether? And what would the owners' explanations look like if they were wrong — would the first-day statements read any differently? They would not. That isn't evidence of anything except the limit of first-day statements, but the limit is the point: nothing published so far can close this question, in either direction.
What would close it is specific. From the government: an attribution statement — or an explicit all-clear — from CISA, the Department of Homeland Security or, for the airline piece, the FAA: agencies that can see across owners instead of grading one homework assignment at a time. From the owners: post-incident reports with the receipts — the named system, the named bug, the failure timeline — that can be checked against the public record. Chase's bug, if it's a bug, has a shape that can be shown. United's failed system has a name. Wells Fargo owes its customers a cause. Until one of those arrives, Monday sits where it sits: three failures, three non-explanations of three different kinds, and a precedent — five quiet years inside the wires — that says the question is a legitimate one to keep asking.
Asking it isn't an accusation. It's the job. The dots on this page are the confirmed ones, the owners' words are printed next to each of them, and the conclusion is deliberately unwritten — because the evidence that could write it hasn't been produced by anyone yet. We'll update this story when it is.
Sources
- Bank outage live coverage: Chase statement and bank responses — USA Today, November 16, 2026
- Wells Fargo acknowledges technical issues — Wells Fargo website, via ABC7, November 16, 2026
- United resolves 'widespread' system outage that delayed some flights — USA Today, November 16, 2026
- PRC State-Sponsored Actors Compromise and Maintain Persistent Access to U.S. Critical Infrastructure — CISA / NSA / FBI joint advisory AA24-038A, February 7, 2024
- United Airlines says tech outage disrupts departures, no signs of cyberattack — Reuters, November 16, 2026
- United Resolves System Outage That Delayed Departures Nationwide — The Modern Leap, November 16, 2026
- Amtrak service halted near Washington after train strikes trespasser — Washington Sun, November 16, 2026
- Cyber-Attack Against Ukrainian Critical Infrastructure — CISA / ICS-CERT alert IR-ALERT-H-16-056-01 (December 23, 2015 attack)
- U.S. Attorney Announces Murder-For-Hire and Related Charges Against IRGC Asset and Two Local Operatives — U.S. Department of Justice, November 8, 2024
- U.S. Treasury's Bessent says he won't be cowed by death threats (CNBC interview) — Reuters, November 13, 2026
- Manhattan U.S. Attorney Announces Charges Against Seven Iranians for Conducting Coordinated Campaign of Cyber Attacks Against U.S. Financial Sector on Behalf of IRGC-Sponsored Entities — U.S. Department of Justice, March 24, 2016
- Exclusive: US intel indicates Iran developing plans to potentially detonate a dirty bomb in Washington, sources say — CNN, November 16, 2026
- Trends in World Military Expenditure, 2024 — SIPRI, April 2025
- Annual Threat Assessment of the U.S. Intelligence Community — Office of the Director of National Intelligence, March 2025
The conversation.
Loading comments…
A little more perspective.
Reading is free. Verified members can comment and reply for $2 a month.