taz calls it “a mistake by an AI”; the Washington Examiner says it was the Pentagon’s “blind faith” in one

What both reports carry. American military planes were already in the air. Armed personnel were preparing to board a Chinese vessel. The intelligence behind the operation said the ship was carrying nuclear weapons, and both reports we hold say that claim was not true. The operation was stopped before anyone boarded.

Where they part. The two reports answer the question of fault in their opening lines, and answer it differently.

	taz	Washington Examiner
Who erred	“ein Fehler einer KI” — a mistake by an AI	the Pentagon’s “blind faith” in it
Who caught it	not said	“more experienced analysts reopened the intelligence”
The nuclear cargo	“angeblich” — allegedly carrying them	its “central claim was false”
Where the account came from	not stated	“according to CNN reporting published Friday”

The German report puts the error inside the software and names no officer, unit or command. The American one moves the actor in its headline: the machine did not nearly start a war, the institution that believed it did.

What is not settled. Neither report we hold carries a response from Washington or from Beijing. Neither names the ship, its owner or its crew, and no one aboard is quoted. Neither dates or places the near-boarding. And both credit the same Friday report by CNN, which we do not hold — so this is two reports resting on one account, not two accounts agreeing.

What this is about. A US military operation against a Chinese merchant ship was prepared and then called off after the intelligence behind it — that the ship was carrying nuclear weapons — was found to be wrong.

How it got here. Stopping and searching a foreign ship at sea is one of the sharpest things one state can do to another short of shooting. Under the law of the sea a warship may not board a foreign-flagged merchant vessel without the flag state’s consent, outside narrow exceptions such as piracy. Since the Proliferation Security Initiative was launched by the United States in 2003, Washington has organised the interdiction of shipments suspected of carrying weapons of mass destruction; China has never joined it. China is one of the five nuclear-weapon states recognised by the 1968 Non-Proliferation Treaty, and its warships and its merchant fleet share the western Pacific with the US Navy, which patrols it continuously. The other thread is newer: American defence intelligence has spent the past decade putting machine-learning tools into the work of sifting imagery, signals and shipping data — software that ranks what an analyst should look at, and sometimes shapes what an analyst believes.

Who is in it. The Pentagon, the headquarters of the US defence establishment, ordered and then stopped the operation; the Washington Examiner, citing CNN, places the intelligence with an analyst at a military special operations command. The vessel is Chinese; neither report we hold names it or its owner. CNN is the outlet that first reported the account, and both reports we hold are built on it.

What is at stake. An armed boarding of a Chinese ship on a false premise would have been a confrontation between two nuclear-armed powers, begun by mistake. If the fault lies in a tool, the answer is a better tool. If it lies in the people who trusted the tool, no tool fixes it — which is precisely what the two reports disagree about.

Where it stands. Neither government has spoken in anything we hold.

The story so far

An AI error 'almost started a war' with China, says The Independent; Frankfurter Allgemeine says it nearly led to action against one ship · 19 Sep 2026

Earlier stories The AI Times filed on this thread.

What is being argued about

Whether the near-boarding was caused by an AI error or by the people who believed its output
What both reports agree on: planes airborne, a boarding party ready, the nuclear-cargo claim false
Two reports, one underlying account — both credit CNN
No response from Washington or Beijing, and no name for the ship or its crew

Two reports, two culpritsSide by side

Outlet	Who erred	Who caught it	Where the account came from	Hedged?
taz	“ein Fehler einer KI” — a mistake by an AI; no officer, unit or command named	not said	not stated	yes — the attack “offenbar” prepared, the nuclear cargo “angeblich” carried
Washington Examiner	“The Pentagon’s blind faith in it”	“more experienced analysts reopened the intelligence behind the operation”	“according to CNN reporting published Friday”	no — the central claim is stated flat as false

Every cell is from the text of the report in that column. “Not said” means the report does not address it.

Sources t taz

Washington Examiner

What we can date, and what we cannotTimeline

not given in either report— no time published US military planes airborne and armed personnel preparing to board a Chinese vessel said to be carrying nuclear weapons; the operation is stoppedneither report we hold gives a date or a place
Friday— no time published CNN reports the account, according to the Washington Examinerwe do not hold this report; both reports we do hold rest on it
Sat 20 Sep 202614:00 time the outlet published Washington Examiner blames the Pentagon’s “blind faith” in AI, not the AI
Sat 20 Sep 202614:08 time the outlet published taz calls the cause “ein Fehler einer KI” and treats AI as a new military riskeight minutes after the other

Times are the ones the outlets published. Neither report dates the incident itself.

Sources

Washington Examiner t taz

The bias in this coverageBias

Two reports, two culprits

Both reports describe the same near-boarding and disagree on who erred: one says a machine made the mistake, the other says the mistake was trusting it. Only one names a human institution. Neither names the ship, its operator or its crew.

2 reports the desk holds on this 1 that name a human institution as having erred 0 that name the vessel or its operator

Nobody from the ship speaks

Both reports are written from the American side of the incident. Neither carries a Chinese official, the shipping company or anyone aboard. One states openly that its account comes from another outlet's reporting, so the desk holds no first-hand version at all.

0 reports quoting a Chinese official 0 reports quoting anyone aboard the vessel 1 reports sourced to another outlet's reporting

Scored without the mastheads. Where the two passes differ, both are kept.

Sources t taz

Washington Examiner

Claims and verdicts

unverified US aircraft were airborne and personnel were preparing to board a Chinese vessel before the operation was stopped.

The Washington Examiner describes the aircraft and the boarding party, and taz describes the same near-attack, but both credit CNN's Friday reporting rather than their own sourcing. Two reports resting on one account are one account.

Checked against 2 sources:

Washington Examiner t taz

disputed An error by an artificial-intelligence system nearly caused a US attack on a Chinese ship.

taz states flatly that the nuclear-cargo claim "war ein Fehler einer KI", while the Washington Examiner's headline rejects that framing — "AI didn't almost trigger a war with China. The Pentagon's blind faith in it did" — and puts the fault on the analysts who trusted the output. The two reports agree the central intelligence claim was false and disagree on what caused the near-boarding.

Checked against 2 sources:

t taz

Washington Examiner

unverified More experienced analysts reopened the intelligence and found its central claim was false.

Only the Washington Examiner reports the re-examination, and it attributes it to CNN. taz says the claim was wrong but does not describe who found it wrong or when.

Checked against 2 sources:

Washington Examiner t taz

Who wrote it

S Sven Hansen taz

What the bias passes found in their reporting

Double hedge Both the planned attack and the nuclear cargo are hedged in one sentence×1

What is hedged the nuclear cargo — the central claim, later found false — is hedged to a single word×1

Who is named as acting The error belongs to a machine; no officer, unit or command is named as having made it×1

Who is named as acting the mistake belongs to the machine; no person or command is named as making it×1

Claims we checked that rest on their reporting

1 disputed2 unverified

From 10 reports we hold under this byline, at taz. Two independent readings, each made without the masthead, a quote attached to every tag. A record, never a rating. Their full record →

B Burak Oktenli Washington Examiner

What the bias passes found in their reporting

Headline makes the argument The headline overturns the machine explanation and names an institution as the party at fault×1

Sourcing at one remove The entire account rests on another outlet's reporting, said so in the text×1

Stated flat The intelligence being false is asserted rather than attributed to anyone who said so×1

Where the account comes from the whole account rests on another outlet's reporting, said so inside the sentence×1

Claims we checked that rest on their reporting

1 disputed2 unverified

From 21 reports we hold under this byline, at Washington Examiner. Two independent readings, each made without the masthead, a quote attached to every tag. A record, never a rating. Their full record →

The bylines on the reports this story is built from. Hover one for what our passes have measured on their reporting.

Coverage, outlet by outlet

t taz —

Fast-Militärvorfall USA-China: KI als neues militärisches Risiko

The only report we hold that puts the error in the machine itself — “Doch das war ein Fehler einer KI” — and the one that hedges the nuclear cargo down to a single word, “angeblich”.

E in mithilfe von KI erstellter, fehlerhafter US-Geheimdienstbericht hätte laut einem Bericht des US-Senders CNN fast dazu geführt, dass inmitten des Irankriegs das US-Militär ein chinesisches Schiff gestürmt und womöglich einen militärischen Konflikt mit der Volksrepublik ausgelöst hätte. Laut vier anonymen Quellen habe das Schiff im Nahen Osten Teile eines Atomwaffenprogramms transportiert. Laut zwei Quellen hätten US-Militär bereits das Entern vorbereitet. Erst dann sei festgestellt worden, dass ein Chatbot falsche Informationen geliefert habe und der Bericht komplett falsch gewesen sei. De

Who is named as acting: The error belongs to a machine; no officer, unit or command is named as having made it — “Doch das war ein Fehler einer KI.”

Double hedge: Both the planned attack and the nuclear cargo are hedged in one sentence — “Das US‑Militär bereitete offenbar einen Angriff auf ein Schiff aus China vor, das angeblich Atomwaffen transportierte.”

Who is named as acting: the mistake belongs to the machine; no person or command is named as making it — “Doch das war ein Fehler einer KI.”

What is hedged: the nuclear cargo — the central claim, later found false — is hedged to a single word — “das angeblich Atomwaffen transportierte”

Read it at taz → https://taz.de/Fast-Militaervorfall-USA-China/!6214745/

Washington Examiner — independent

AI didn’t almost trigger a war with China. The Pentagon’s blind faith in it did

The only report we hold that names a human party at fault, that describes who caught the error — “more experienced analysts reopened the intelligence” — and that says in its own text where the account came from.

American military planes were already in the air. Armed personnel were preparing for a possible boarding of a Chinese vessel. Then, according to CNN reporting published Friday, more experienced analysts reopened the intelligence behind the operation and found that its central claim was false. CNN reported that an analyst with a military special operations command used an AI chatbot to interpret classified and open-source information about a Chinese ship in the Middle East. The chatbot identified the cargo as a component of a nuclear weapons program. The assessment circulated, and U.S. forces b

Headline makes the argument: The headline overturns the machine explanation and names an institution as the party at fault — “AI didn’t almost trigger a war with China. The Pentagon’s blind faith in it did”

Sourcing at one remove: The entire account rests on another outlet's reporting, said so in the text — “according to CNN reporting published Friday”

Who is named as acting: The boarding party appears without a nation, service or unit attached — “Armed personnel were preparing for a possible boarding of a Chinese vessel.”

Stated flat: The intelligence being false is asserted rather than attributed to anyone who said so — “more experienced analysts reopened the intelligence behind the operation and found that its central claim was false”

Who is named as acting: the headline moves the actor from the machine to the institution that trusted it — “AI didn’t almost trigger a war with China. The Pentagon’s blind faith in it did”

Who is named as acting: the boarding party has no nationality, unit or commander; the planes do — “Armed personnel were preparing for a possible boarding of a Chinese vessel.”

Where the account comes from: the whole account rests on another outlet's reporting, said so inside the sentence — “according to CNN reporting published Friday”

Read it at Washington Examiner → https://www.washingtonexaminer.com/op-eds/4735008/pentagon-ai-intellige…

Checked, didn’t cover

International Criminal Court

UN News

Silence is recorded, not inferred — each of these was read when we built this story and carried nothing on this event.

We could not read these

International Court of Justice

These are must-cite outlets for this desk that we did not manage to read when we built this story. That is a gap in our reach, not a silence on their part, and it is not evidence of anything about this story.

Read another desk

Sections

Registers

Desks

taz calls it “a mistake by an AI”; the Washington Examiner says it was the Pentagon’s “blind faith” in one

The story so far

What is being argued about

Two reports, two culprits

Nobody from the ship speaks

Claims and verdicts

Who wrote it

Coverage, outlet by outlet