How causes are judged

The first story about an outage is often wrong. In one week in November 2026, early reports blamed hackers for the US ground stop and AI for a cloud outage; neither held up as first told. So no cause here rests on one account.

3+1: FOUR SIDES OF EVERY CAUSE

1 OFFICIALWhat the agency or company says, who said it, and whether they show evidence (named forensics, logs, an investigator) or only assert it. "No indication of an attack" with nothing behind it is shown as a claim.
2 INSIDERSThe people who run the system: workers and their unions, operators, contractors.
3 OUTSIDE EYESMeasurements nobody in charge controls: internet routing and reachability, flight traffic, grid data, independent researchers, investigators such as the NTSB, inspectors general or the GAO, and reporters' own sources.
+1 THE CHECKWhere the accounts disagree. The timing: who asked for what, before and after. The inside and supply-chain questions: who made, shipped, installed and set up the part, who had access, whether the failing sites share a batch or a crew, and whether the logs were kept and shown.
Then the OPEN QUESTIONS: what is still unanswered, including who gains. These stay questions; they never decide a verdict.

VERDICTS

CONFIRMED ATTACKthe attack is established by the victim, investigators or the attacker.
OPENthe accounts conflict or the evidence that would settle it has not been shown.
CLAIMED ACCIDENTthe company or agency calls it an accident; named in brackets.
UNCHECKEDreported only by a round-up so far; not yet checked against the original.
NO CAUSE GIVENa live measurement of what is out; no one has said why yet.
DEBUNKEDa rumour shown only because it has been disproved.

A cause is SETTLED only when the outside measurements agree with it.

WHY THE INSIDE QUESTIONS MATTER

A planted or swapped part looks exactly like an accident from outside. On record:

THE PATTERN

Along the bottom of the globe every kind of outage runs on one timeline. It shows separate systems failing close together, and one target hit again and wider (one vendor, one country, then many). A cluster is shown as a cluster; it is called one actor only when the sources show it.

WORKED EXAMPLE: THE 13 NOVEMBER GROUND STOP

1 Official: a new router on copper-to-fiber work flooded the FAA's network; Homeland Security saw "no indication" of an attack, with no forensics named.
2 Insiders: the FAA's own administrator said the router story "did not serve as a full explanation of the root cause"; controllers reportedly "don't buy" it; their union points to years of underfunding.
3 Outside eyes: no independent investigation or network data published yet.
+1 The check: officials gave differing causes; $30B was asked on 12 November, the outage came on the 13th, and the ask became $50B with no itemised list.
Verdict: OPEN See it on the globe