HomeArtificial IntelligenceIs the United States Under Cyberattack? What Two Weeks of Attacks and...

Is the United States Under Cyberattack? What Two Weeks of Attacks and Outages Actually Reveal

A Review of Publicly Available Evidence From November 5–18, 2026

The United States is experiencing real cyberattacks against businesses, universities, public services, and technology providers. Some have disrupted operations; others have exposed information or delivered malicious software through trusted suppliers. However, the evidence reviewed through November 18, 2026, does not establish that the high-profile outages constitute a single coordinated attack on the country. The Federal Aviation Administration, Microsoft, JPMorgan Chase, Apple, and the National Security Agency have reported technical or maintenance-related explanations for their respective disruptions.

Responsibility is also more complicated than naming one adversary. The record includes financially motivated criminals, attackers whose identities remain unknown, and separately documented state-sponsored operations. A November 12 FBI advisory directly attributed a cryptocurrency theft campaign to North Korean state-sponsored hackers. That attribution does not connect North Korea to the airline, banking, email, or intelligence-network outages.

The strongest assessment is specific: American organizations are under sustained cyberattack, but a nationwide coordinated cyber offensive explaining the November outage cluster has not been demonstrated publicly.

This review distinguishes events occurring during the 14-day period from older attacks disclosed or updated during it. It prioritizes official incident notices, government advisories, provider reports, and reporting that includes statements from affected organizations. It is a review of significant publicly documented events, not a complete census of every intrusion or service interruption.

The Major Outages: What Is Actually Known

Several widely used systems failed within days of one another. Their concentration understandably raised concern, but the available explanations differ. The table summarizes the principal disruptions and the evidence concerning their causes.

DateOrganization or ServiceReported DisruptionEvidence Concerning the Cause
November 12–14, 2026Microsoft 365International disruption of email and functions dependent on Exchange Online.Technical failures; no indication of malicious activity.
November 13, 2026Federal Aviation AdministrationNetwork instability prompted a nationwide ground stop.Excessive traffic from two sites; no indication of a cyberattack.
November 13–14, 2026WorkdayWD1 production outage affected business applications and automated processes.A disaster-recovery environment problem left production data inconsistent.
November 15, 2026Amazon Web ServicesNorthern Virginia region experienced authentication, signup, and connectivity problems.AWS documented recovery; no attacker or definitive root cause identified.
November 16, 2026JPMorgan ChaseSome transactions processed slowly.Software bug in on-premises systems; no evidence of cyber-related cause.
November 16, 2026United AirlinesDepartures delayed; airborne flights and United Express unaffected.Technology problem; no indication of a cybersecurity cause.
November 17, 2026AppleServices including iCloud and Apple Pay experienced disruption.Apple attributed the outage to an internal update, excluding a cyberattack.
November 17, 2026National Security AgencyInternal network degradation; essential national-security capabilities remained operational.NSA attributed the problem to scheduled maintenance.

These classifications reflect what the organizations had established publicly by November 18, 2026. They do not imply that every investigation is complete. An unexplained operational incident should remain classified as unresolved rather than automatically becoming either a confirmed attack or a confirmed nonmalicious failure.

Microsoft’s preliminary explanation provides more detail than a generic reference to technical difficulties. Its preliminary incident report describes a latent defect that could overwhelm networking gateway servers under particular load conditions. The change associated with that defect had been deployed months earlier without observable disruption. A subsequent traffic shift increased load on the affected infrastructure. The report distinguishes the underlying incident window from the broader customer recovery process and remains preliminary.

The FAA’s network traffic surge also requires careful interpretation. A flood of connection attempts can resemble an attack to an outside observer, but excessive traffic can also arise from equipment behavior or configuration problems. The FAA said isolating two locations stabilized its network and that investigation continued. Air traffic resumed at 8:15 p.m. Eastern on November 13. Its published assessment explicitly stated that available evidence did not indicate a cyberattack or outside malicious actor. The term “flooding” alone does not establish a denial-of-service attack.

Banking disruption can spread without multiple banks being independently attacked. JPMorgan said other institutions could encounter delays when they were counterparties to its transactions. Consequently, complaints involving several banks would not, by themselves, prove several separate compromises.

The NSA incident understandably attracted particular attention. Nevertheless, the agency’s own statement described maintenance-related degradation, not a successful hostile intrusion. Public information remained limited on November 18; assigning responsibility to a foreign government would go beyond the evidence.

Confirmed Attacks and Security Incidents During the Same Period

The absence of evidence connecting the major outages does not diminish the attacks that organizations actually confirmed. The following incidents distinguish documented operational effects from unresolved attribution.

No.OrganizationDevelopment During the Review PeriodAttribution or Remaining Uncertainty
1Vanderbilt UniversityNovember 9 incident restricted devices, systems, and cloud services.BYOD claimed responsibility; university updates did not confirm attribution.
2DonorPerfectRestricted platform access after detecting unauthorized activity in November.Security incident confirmed; responsible group not publicly identified.
3Monadnock Community HospitalNovember 16 notice confirmed a cyber incident caused its server outage.Data impact under investigation; perpetrator not identified.
4South Portland, MaineLate November 14 discovery led to municipal disruption and office closures.Preliminary assessment indicated cybercriminal activity; no named actor established.
5Optimizely Web ExperimentationNovember 16 signup vulnerability exploitation inserted malicious JavaScript into customer projects.Compromise confirmed; no responsible state or named criminal group identified.
6Expedia GroupNovember 12 confirmation: attackers accessed systems through a provider.n0n claimed data theft; claimed scale unvalidated; vendor unidentified.
7InfobloxNovember 16–17 disclosure concerned unauthorized access through a data-migration vendor.No indication products, cloud services, or software release processes were involved.
8Charleston County, South CarolinaUnusual activity detected November 16; data removal confirmed.Investigation remained open; normal public services continued.
9Perry Local Schools, OhioCybersecurity incident disrupted technology and closed school buildings November 16.Attack method, responsible actor, and data impact not disclosed.
10University of Chicago Computer Science DepartmentNovember 7 reporting described attacks, account restrictions, and departmental service shutdowns.Staff reported no compromised user data; perpetrator not identified.

These incidents demonstrate real harm across education, healthcare, local government, fundraising, and commercial technology. They do not establish a common perpetrator.

At Vanderbilt, the university’s November 17 update said OneDrive, SharePoint, and Teams remained disabled, with restoration expected to take weeks. It described continuing work to support research and essential university functions. Significantly, Vanderbilt had explicitly stated that the November 12 global Microsoft outage was unrelated to its own cybersecurity incident. The same institution could suffer from a confirmed attack and an independent supplier outage at the same time.

Monadnock illustrates the practical consequences of a hospital technology disruption. Its November 17 operational notice said the emergency department remained open, but ambulance diversion continued. Some elective major surgeries were canceled through November 23, while outpatient radiology and laboratory services remained unavailable. A locally concentrated incident can impose substantial costs on patients even without becoming part of a nationwide campaign.

DonorPerfect’s incident interrupted access to software used by nonprofit organizations during an important fundraising period. Its president described taking the platform offline to protect information while specialists investigated. The public evidence establishes unauthorized activity and operational disruption; it does not establish that a government directed the attack.

Security-Related Shutdowns Are Not Always Confirmed Breaches

Two additional developments warrant inclusion without overstating what is known.

Court portals using Journal Technologies experienced a shared security problem. Alameda County’s November 17 update said the vendor was implementing security updates and that its public portal and the JTI-specific filing service remained offline. Marin County’s court notice described a potential cybersecurity threat and precautionary suspension of its portals. These notices support classifying the event as a security-related disruption under investigation. They do not establish that every affected court was separately breached or that court records were stolen.

The Los Angeles Homeless Services Authority took systems offline while investigating a potential security issue. Employees received shutdown instructions on November 6. In its November 9 statement, the authority described the action as precautionary and said outside specialists were investigating. Officials had not established whether sensitive information had been accessed. Reporting this as a confirmed ransomware attack or a proven theft of client records would exceed the available facts.

A defensive shutdown can be an appropriate containment measure. Its visible effects may resemble those of an attacker disabling the systems directly, but those are different events.

Older Attacks Produced New Disclosures and Continuing Disruption

The distinction between an attack date and a disclosure date materially changes the picture.

Queens Public Library continued recovering during the review period, but the outage began November 4, just outside it. November 10 reporting described continuing restrictions on digital services following a ransomware attack. The perpetrators’ identities and any ransom demand remained undisclosed. This belongs in an account of the period’s disruption, but should not be counted as an attack that began after November 5.

Fireworks AI published security updates on November 9 and 17 about two earlier incidents. Its November 17 bulletin separated activity spanning August 3–October 7 from another incident spanning October 23–30. The company reported exposure affecting account information and, for a smaller set of customers, datasets, models, or secrets. Its earlier update described evidence pointing toward financially motivated abuse of AI resources. These are substantial disclosures, but they do not establish new November intrusions.

International companies also reported compromises relevant to American customers and supply chains. TSMC’s November 12 filing disclosed unauthorized publication of employee information matching records held by a cloud provider. It said there was no evidence its own systems had been compromised and distinguished the exposed information from operations and customer data. The reproduced regulatory filing does not establish disruption of semiconductor production.

Maersk confirmed unauthorized access to an inherited legacy system, while the WhiteArmor group claimed responsibility and demanded payment. The company said other systems and customer operations were unaffected, as summarized in Check Point’s November 9–15 threat report. This is relevant international context, not evidence that American transportation infrastructure was being disabled by the same attacker.

The Optimizely Attack Shows How One Compromise Can Reach Many Organizations

Among the period’s confirmed attacks, Optimizely provides a particularly clear example of how malicious activity can spread through a shared service.

Websites use Optimizely to test different page designs and user experiences. Its November 17 incident report said attackers exploited a flaw in the account-signup process to impersonate users and modify JavaScript in customer projects. The company said this did not require stolen customer passwords. It reported closing the vulnerability and removing injected code on November 16.

Report URI’s investigation identified infected snippets associated with 178 organizations. Its findings included payment-card skimming, cryptocurrency address substitution, and attempted exploitation of older Apple devices. The researchers identified affected web properties across several sectors and countries. The number of organizations with infected snippets should not be interpreted as a verified count of successfully compromised visitors.

This incident supports a specific connection between affected websites: they received malicious code through the same service. It does not connect that campaign to the FAA ground stop, the NSA maintenance incident, or JPMorgan’s transaction delays.

The distinction matters because a supplier compromise can generate dozens of recognizable names in headlines. A long list of affected brands may represent one shared intrusion path rather than dozens of independently coordinated attacks.

Who Is Responsible?

The answer depends on which incident or campaign is being discussed. Attribution should distinguish an attacker’s claim, a researcher’s assessment, a victim’s confirmation, and a government’s formal finding.

Financially motivated criminals are implicated in several cases, but identities often remain unresolved. BYOD claimed the Vanderbilt attack. Reporting on the BYOD claim included a ransom note and other material supporting that connection, while noting that Vanderbilt had not officially confirmed the group’s involvement. Expedia’s confirmed supplier compromise was accompanied by claims from n0n, but the company did not validate the claimed data volumes. Neither case provides a sound basis for assigning responsibility to a foreign government.

North Korean state-sponsored hackers were explicitly identified in an advisory issued during the period. The FBI’s November 12 FaceStealer warning described attackers using hijacked accounts, fake video meetings, and recordings of previous victims to deceive new targets. The bureau attributed the campaign to North Korean state-sponsored hackers and reported more than $400 million in cryptocurrency theft over 18 months. That figure covers the longer campaign – not the 14 days reviewed here – and does not establish any connection to the major outages.

Chinese government-linked activity is documented in separate, earlier official reporting. An October 8 joint advisory described actors combining automated tools and direct exploitation to steal sensitive information, including from U.S. infrastructure sectors. This establishes relevant background about an existing threat. It is outside the review window and cannot be used as proof that China caused November’s service failures.

Russian and Iranian activity also appears in earlier government advisories. A July advisory on LAUNDRY BEAR attributed email-focused espionage to Russian state-supported actors. A separate advisory, updated July 22, addressed Iranian-affiliated exploitation of industrial controllers in U.S. infrastructure. These records justify taking those threats seriously, but they do not attribute the incidents in this article to those governments.

For many of the November incidents, the responsible actor remains publicly unknown. That is an evidentiary limit, not a reason to substitute the name of a familiar adversary.

Is Artificial Intelligence Behind the Attacks?

Evidence published during the review period supports a growing role for AI in malicious activity.

The FBI’s November 13 advisory on ARTEX described an autonomous penetration-testing framework being used by attackers to automate phases of intrusions. It reported malicious use since at least August and described some operations progressing to privileged access in less than three hours. The advisory concerns multiple actors and sectors, rather than identifying one AI-directed national campaign.

Google Threat Intelligence Group’s November 16 report on agentic threats similarly described attackers using agentic tools, automated vulnerability analysis, and stolen AI resources. Its findings covered activity over the preceding quarter. They provide evidence of changes in attacker capabilities, but they are not a count of attacks initiated during the last two weeks.

These findings support concern about faster exploitation and shorter response windows. They do not establish that AI caused the Microsoft, FAA, banking, or NSA outages. Microsoft specifically rejected an AI-agent explanation for its disruption.

AI is also not a sufficient answer to “by whom?” A tool can accelerate an operation without identifying the person, criminal organization, or government directing it.

Other Warnings Published During the Period

The government advisories issued during these two weeks point to risks extending beyond the most visible outages.

On November 5, U.S. agencies warned about malicious activity targeting industrial refrigeration and cold-chain monitoring systems internationally. The advisory described risks to temperature controls used in food, healthcare, and transportation. It urged operators to reduce internet exposure, secure remote access, and strengthen authentication. It did not establish a nationwide American refrigeration failure or identify a single perpetrator responsible for the broader outage cluster.

On November 16, the FBI issued a Gootloader warning concerning malicious search results and compromised websites used to gain access to organizations. The advisory described a multiyear campaign, including targeting of U.S. legal professionals. Its figures for infected websites and ransomware payments were historical campaign findings, not totals for November 5–18.

Together, these warnings show why attention should extend beyond services that visibly go offline. Theft, espionage, credential compromise, and malicious code delivery may occur while a service continues functioning.

Does the Timing Suggest Coordination?

Timing can justify investigation. It cannot establish coordination on its own.

A persuasive connection would require evidence such as shared attacker-controlled infrastructure, distinctive malicious code, a common compromised account or supplier, matching forensic findings, or a supported attribution from investigators. The public record reviewed contains a demonstrated shared delivery mechanism for the Optimizely campaign and a shared vendor issue affecting court portals. It does not establish comparable links among the major airline, banking, email, and intelligence-network outages.

Nor does this review provide a defensible statistical basis for declaring a new national attack surge. Such a finding would require a consistent historical baseline, comparable reporting coverage, and deduplication of incidents and downstream victims. Combining fresh disclosures, older attacks, precautionary shutdowns, and technical failures would inflate any apparent count.

There are also limits to official assurances. Statements issued early in an investigation may change, and some evidence may remain confidential. An organization’s initial denial should be assessed alongside later technical findings. Conversely, a lack of complete public detail is not affirmative evidence of a hidden attack.

The appropriate position is to preserve uncertainty where it exists while giving greater weight to documented mechanisms than to the coincidence of dates.

What the Evidence Means for the United States

The period reveals two simultaneous problems.

The first is deliberate malicious activity: unauthorized access, extortion, supplier compromise, data theft, and attacks on public-facing services. The second is operational fragility: failures in heavily shared systems can interrupt travel, payments, communications, and government work across a large area.

Those problems can overlap. A single organization can be responding to an intrusion while also losing access to a cloud supplier experiencing an unrelated outage. Vanderbilt explicitly documented that distinction. Shared vendors can also transmit either a technical failure or malicious code to many customers, although the mechanisms and responsibilities differ.

The public evidence available through November 18, 2026, supports clear answers to the article’s central questions:

  1. Is the United States experiencing cyberattacks? Yes. Confirmed incidents affected American institutions during the review period.
  2. Do the major November outages establish one coordinated attack on the country? No. The reviewed evidence does not establish that connection, and several operators have published technical explanations.
  3. Who is responsible for the confirmed attacks? Responsibility differs by incident. Some criminal groups have claimed attacks, many perpetrators remain unidentified, and the FBI has separately attributed a longer-running theft campaign to North Korean state-sponsored hackers.
  4. Can China, Russia, Iran, North Korea, or an autonomous AI system be named as the cause of the entire outage cluster? Not on the evidence reviewed.

Cyberattacks are causing measurable harm, and major technical failures can produce widespread disruption without an attacker. Treating every outage as hostile obscures the failures that need to be corrected. Treating the entire period as ordinary technical trouble overlooks the organizations that were demonstrably attacked.

YOU MIGHT LIKE

WEEKLY NEWSLETTER

Subscribe to our weekly newsletter. Sent every Monday morning. Quickly scan summaries of all articles published in the previous week.

Most Popular

Featured

FAST FACTS