Series Failures in the United States: Change Management Instead of Attack, According to Incident Reports

In just one week, the United States experienced a cascade of outages affecting Microsoft 365, air traffic control, banking systems, and the NSA’s internal network. Reports from Microsoft and the Federal Aviation Administration (FAA) have attributed these disruptions to poorly compartmentalized infrastructure resulting from technical changes. This situation serves as a critical lesson in change management for IT departments, especially as the number of patches continues to rise.

Between November 12 and 17, a series of outages impacted various digital services and American infrastructure. Many Microsoft 365 clients faced outages from November 12 to 13, and on November 13, the FAA grounded all flights in the country for over four hours. Additional incidents arose in banking, air transport, and even at Apple in the days following, yet no technical report matching the detail of the earlier ones had been published by November 18.

The simultaneous nature of these outages fueled public speculation about a coordinated attack or an uncontrolled artificial intelligence system. However, the only two organizations that released detailed analyses, Microsoft and the FAA, pointed to routine mechanisms at play rather than any nefarious activity.

Two Reports, One Common Cascade Mechanism

Microsoft’s preliminary report cites three interrelated factors as contributors to the outage. A latent defect combined with a contention for locks that strained server CPU and memory resources was introduced through a network gateway patch deployed in July and had gone unnoticed until that week. About fifteen hours before the outage, Teams traffic was shifted to the gateway also serving Exchange Online. Compounding the issue, roughly 14% of the servers on that gateway had been down for another reason since the previous day. Load testing had failed to reveal the defect because labs utilized less powerful hardware configurations.

The FAA’s statements echo this sequential pattern. A surge of data, with approximately 1,000 connection attempts every four to five seconds, overwhelmed its FAA Telecommunications Infrastructure (FTI), destabilizing aging equipment. This resulted in complications across 17 of the 22 air traffic control centers. The FAA traced the issue back to two locations: a radar system in Amarillo, Texas, and a radio site in El Dorado, Arkansas, where their contractor L3Harris was deploying a new router.

Technical Changes at the Heart of the Analyzed Incidents

In both cases, outages stemmed from technical changes — whether a patch, traffic rerouting, or equipment installation — rather than an external intrusion. These disruptions escalated due to a lack of effective containment within the architecture. According to Microsoft, the outage began during peak traffic hours in the U.S. and spread throughout its global infrastructure due to inadequate compartmentalization. For the FAA, just two sites were enough to destabilize much of its operational control.

As a result, Microsoft suspended all planned deployments and changes on November 12 at 19:00 UTC, subsequently retracting the faulty change itself. The FAA disabled the problematic function on the affected Juniper routers, halted their installation, and announced it would reassess daytime maintenance practices after routine replacements were conducted at remote sites. The Department of Transportation also noted that these routers were purchased in 2001 and deemed obsolete as far back as 2006.

No Evidence of Attack or AI Involvement, According to Primary Sources

None of the affected sources implicated AI in their reports. Microsoft described the incident as a “technical problem unrelated to cybersecurity, AI, or agents.” The FAA confirmed that there were no signs of a cyberattack or external malicious actor and stated that its new SMART system, an AI-based data aggregator, was not linked to the outage. Claims of direct AI involvement are unsupported by any published evidence.

However, a coincidence has emerged: four of the impacted organizations — Microsoft, Amazon Web Services, Apple, and JPMorgan Chase — are founding partners of Project Glasswing, an initiative by Anthropic that provides access to Claude Mythos Preview, its vulnerability discovery model. These organizations represent the largest tech and financial companies in the U.S.; notably, the FAA, United Airlines, and the NSA were not included in this group. Moreover, Anthropic expanded the initiative on June 2 to around 150 organizations across more than fifteen countries and publicly offers Claude Fable, which is governed by cybersecurity safeguards. Thus, this program does not serve as a common thread for the outages.

The Patch Frequency Theory Remains to Be Verified

An indirect theory remains open but unproven: the role of models that identify vulnerabilities and prompt numerous patches, with each patch representing a change. Anthropic reports that Glasswing partners identified over 10,000 critical vulnerabilities. Google’s threat intelligence team found that vulnerability disclosures nearly doubled from January to August 2026. In France, CERT-FR issued its CERTFR-2026-AVI-1555 advisory on November 17, with numbering reaching around 1,000 on that same date last year. Furthermore, as early as May 7, Senate Democratic leader Charles Schumer urged the Department of Homeland Security to develop a plan for rapid patch deployment.

This theory has its limitations; Microsoft’s patch was from July, and the FAA was replacing a router during network line conversions, with no established link to any AI-discovered vulnerabilities. Importantly, while the pressure for patches is global, published incidents have mainly impacted domestic operators, apart from Microsoft’s globally affecting outage. In the editorial team’s assessment, the aging of certain public infrastructures and the concentration of control plans among major providers explain the geographical occurrences more effectively than solely the pace of changes.

Change Containment: A Priority for IT Departments

For client companies, the two reports shift the focus from threats of attacks to the management of change, both internally and externally with suppliers. Testing a patch on hardware less powerful than production has proven insufficient for Microsoft, and a daytime intervention at a distant site was enough to disrupt U.S. air traffic control.

Demanding post-incident reports from suppliers, detailing the portion of their infrastructure affected by a single deployment, is crucial for assessing how effectively their containment mechanisms perform. Microsoft’s final report, expected on November 20, will reveal whether these barriers have been strengthened.

Scroll to Top