AI false report nearly led US to attack a Chinese cargo ship

newsSep 18, 2026392,282

This spring a US intelligence report that used artificial intelligence falsely flagged a Chinese cargo ship in the Middle East as carrying components of a nuclear weapons program, prompting plans to intercept the vessel, CNN reports. Four sources said US military personnel prepared to board the ship and military planes were airborne before officials discovered the analyst had relied on a chatbot that had inaccurately identified the cargo. The chatbot fused open-source information with classified signals intelligence from government holdings, and the analyst then used AI again to package and disseminate the standard intelligence report. One source called the report entirely false and said the episode almost started a war because any US action against a Chinese vessel risked spiraling into armed conflict. CNN says it was unclear whether the chatbot was commercially available or a government tool and that US Special Operations Command Pacific originated some of the reporting the analyst queried. The incident came as the Pentagon and other agencies accelerate AI adoption, and it underscores the risks of using relatively poorly understood AI for targeting decisions in wartime.

public health guy

@publichealthguy1.bsky.social

so we just nearly started a war with china (a nuclear state) because of genAI use

The intelligence report, circulated across the US military this spring in the midst of the war with Iran, immediately set off alarm bells: A Chinese ship in the Middle East was transporting components of a nuclear weapons program.
The US military swung into action with plans to intercept the vessel, according to four sources familiar with the episode.
According to two of the sources, armed members of the US military were preparing to board the ship. Military planes were in the air, one of those sources and another source familiar with the incident said.
It was only just before the planned operation that officials dug deeper into the report put together by a special operations command analyst and found it had been generated with the help of artificial intelligence (AI) — and that a chatbot the analyst had used inaccurately identified the material the ship was carrying. CNN was not able to learn what the misidentified cargo was.
The report, according to one of the sources, was "entirely false." But it also "almost started a war," the source said. Any US operation against a Chinese vessel could have risked spiraling into an armed conflict between the two nations.

69521d ago

public health guy143

CNN really downplays the seriousness of this incident in the headline, for whatever reason

Tinker Tailor Super Fly22

wait, but am i reading here that if there WERE nuclear components on the Chinese ship, we *would* have boarded them?!

Redshift ❌👑5

Remember when Elon/DOGE said firing all these federal workers with decades of expertise was no problem because we could replace them with AI? How'd that work out?

Jessica5

This is how AI will kill us?

Ben Schwartz5

fighting iran and china at the same time is a seagal movie.

kshaughn75.bsky.social4

You read something like this, and then think about the people currently in charge, and the full realization of how fucked we truly are suddenly hits you. We're basically wholly reliant on Stanislav Petrovs to save all our asses now--which is not a viable situation.

Quoting this

Mike Drucker385

On one hand, AI might cause nuclear annihilation in the dumbest way possible. But on the other hand, it is semi-almost-useful at making lists. Gotta take the bad with the good!

whet moser56

The Dumb of All Fears

phil48

i see the chatbots have graduated from "you should kill yourself" to "you should kill everyone on the planet"

🌷 Fidget/Vesper 🌷16

My understanding of AI, as a software engineer, is that a complicated statistical model won't suddenly get sentient and decide to kill us; but people who consider AI a magic authority will use it irresponsibly in high-stakes decisions, and those people *may* kill us

Chris Plante15

There’s a scene in Fail Safe where the POTUS sacrifices NYC as a make good for the accidental bombing of Moscow. I think Trump would be like, “why not take all of California too?”

Dr. SkySkull12

It's wild that people in power are like "hey, let's put this global life and death decision in the hands of a black box that is infamous for making spectacular mistakes"

Dan the Gnome10

WHAT THE ACTUAL FUCK. We're going to kill ourselves with AI long before it has the capability to kill us.

MayaScientist8

Turns out there is like one AI doomsday scenario that isn't completely far fetched but it's not one the AI companies seem to care about for some reason.

Der @KleineMaulwurf7

This is honestly the only way that AI is gonna wipe out humanity. By having humans put critical decisions to the stochastic parrot.

Cathy Kimport5

If AI actually ends the world it’s going to be because of boneheaded shit like this, not some tech bro fever dream

Max Kennerly

@maxkennerly.bsky.social

Lot of issues here but one part that sticks out in light of the almost-attacked-a-Chinese-ship story today: people with very important jobs where lives are at stake have bought the hype and outsourced their whole brains to AI. We're going to see more and more of this. www.cnn.com/2026/09/18/p...

In the past year, the Defense Department has rapidly made Maven a cornerstone of the US military apparatus. Several former senior US military officials now work for Palantir, including some with high-level security clearances in Centcom’s Tampa headquarters. Their role is to keep the software functioning inside secure military environments. Some Centcom personnel expected Maven to flag stale information or inconsistencies in the underlying intelligence assembled for potential targets, the officials said. It’s not clear why they had such expectations.

58721d ago

Greg Greene (he/him/his)151

That uncanny feeling when Sonnet 4.6 evaluates the core cognitive failure uncovered in both stories with more acuity than the dingbats at the Pentagon:

Randy Herman23

I love the sentence "it is not clear why they had such expectations"

public health guy20

yeah, this problem goes a lot deeper than just hegsgeth

JLoHayLo9

Meanwhile just today Amazon’s chatbot didn’t know what day of the week or date it was in my interaction with it

tbluvoter❌👑6

Who are the AI goons infiltrating our military & our govt actually working for? Not us. A former senior US official familiar w/ military AI tools told the publication, “The internal tools are mostly just copies of the commercial stuff wearing lipstick” www.the-independent.com/news/world/a...

Michael Planey3

This ⬆️ is the biggest issue and you are absolutely correct. Everyone concerned about “AI is going to end the world“ aren’t wrong, but they’re worried about the wrong thing. If AI induces an extinction level of event, it’s going to be because of the failure of the people who create and use it.

Quoting this

David Kaye128

it's really essential to underline that the use of failing or problematic software tools/AI tech IN NO WAY ABSOLVES THE ACTORS OF CRIMINAL CULPABILITY.

Josh Fruhlinger107

so much less worried about "what if claude goes rogue and hacks an bioweapons lab" than i am about "what if claude tells a one-star something wildly incorrect and he believes it and starts a war"

Brendan Nyhan58

Makes me think of Stanislav Petrov, one of the greatest heroes in history. We're going to need to build systems that don't rely on people like him to stop a rush to judgment.

Blayne Haggart32

Much, much more. In Canada, as @ntusikov.bsky.social and I point out in a recent Policy Options piece, Mark Carney's foolish all-in bet on AI, combined with his civil service job cuts, is guaranteed to put lives at risk. People will die.

oh jeez2

"it's not clear why they had such expectations," is such a delicious bit of criticism.

Katelyn Burns

@katelynburns.com

AI hallucinated a report that a Chinese ship was transporting nuclear material to Iran and the US military scrambled and nearly boarded it. We were apparently minutes away from a major war because of an AI hallucination. www.cnn.com/2026/09/18/p...

34921d ago

s. e. smith20

Great system we have here

Irritable Yet Soporific9

Not the only time this Spring. On 7 April, Trump threatened to slaughter everyone in Iran. 4 hours later, a nuclear command jet took off from Andrews AFB and flew at a speed slow enough to trail a long-wire antenna to talk to ICBM bases and missile subs for almost 2 hours.

Amanda E 🇨🇦🇪🇺8

AI execs: That's okay, these are just teething problems, our models are getting better all the time, anyway give us more money /s

Rewa3

The raging lunatic ceo promised us they needed all that contracts for the US to win 21st century. Instead they’re profiteering, and killing school girls….the us missles are depleted in the middle east, chinese gps system are helping iran with their cheap drones…. so where’s investment $$

Gideon1

My fucking nervous system can’t take this shit

Quoting this

Euan Thomson37

Canada should train thousands of people on how to use this and push it into every public sector

Katelyn Burns37

Subscribe to @theflytrapmedia.com, we don't use AI and we won't start any wars.

Gus Squawks22

Is this why they just banned CNN from the White House (not a thing you can do btw)

Jake Williams19

I want to be surprised. I am not surprised. This was absolutely foreseeable. That there wasn't a system of checks here to uncover this before we were nearly at war is a failure of leadership.

suzette sommer15

THIS IS INFURIATING. How stupid ARE THESE PEOPLE!!!!????

Mz. April Daniels ⚧13

We almost started WW3 because a chatbot told us to. Iran + China + a proxy war with Russia in Ukraine = WW3. On top of that, the shitheel still wants to invade Greenland, which would trigger NATO's Article 5 against us. We are *so* close to the abyss, and the news won't cover it that way.

Shannon E French13

🚨 How is this not a bigger story? Seems pretty IMPORTANT to me…

Christy Somos9

I keep coming back to this story because the fact that no one is screaming from the rooftops about it in govt and immediately doubling down on AI regulations is mind-boggling

Andrew [Retired from Sad]7

You guys ever win an argument three years after it happened?

fuck fascism (she/her 🏳️‍⚧️)5

i dont respect you if you use generative ai or support the military in any capacity. straight up

Nicholas Grossman

@nicholasgrossman.bsky.social

When the US bombed an Iranian school, killing over 100 kids, it involved over-reliance on an AI targeting system called "Maven," which based it on outdated information. I called it in @foreignpolicy.com in April Now confirmed in @bloomberg.com report of Pentagon investigation

Keep Humans in the Loop
AI has a place in military targeting—but it needs safeguards.

By Nicholas Grossman, a professor of international relations at the University of Illinois, editor of Arc Digital, and the author of Drones and Terrorism.

Pictured: A woman passes an art installation featuring portraits of some of the school girls killed in an apparent U.S. strike in February, in Tehran.on April 12. MAJID SAEEDI/GETTY IMAGES)

https://foreignpolicy.com/2026/04/14/ai-targeting-iran-school-airstrikes-pentagon-anthropic/

21921d ago

Nicholas Grossman26

From Ben Bartenstein and Krishna Karra's reporting in @bloomberg.com www.bloomberg.com/graphics/202...

JQP2

If bombing a school of little girls out of pure negligence isn't an impeachable offense, I don't know what it

Quoting this

Eric Blair22

Maven is built on Anthropic’s Claude. last January and February, Anthropic beefed with Trump and Hegseth about not using their AI to make autonomous killer drones. Meanwhile their AI was being used to target bombings, including to a girls‘ school.

Ambassador Ken Fairfax (ret)

@portlandken.bsky.social

If the US had attacked a Chinese ship for no reason (and there was none), it easily could have started a major war with a nuclear weapons country. This is one of the many ways AI could doom humanity, particularly with idiots like Trump and Hegseth in charge.

5921d ago

🦋💔🇺🇸SEB🇺🇸💔🦋

@sebastiangarof5.bsky.social

An AI-generated intelligence report triggered preparations for a US operation, but it was caught before any war started. CNN’s exclusive says a chatbot wrongly flagged a Chinese ship as carrying nuclear weapons components during the Iran war. 🔄 www.cnn.com/2026/09/18/p...

2521d ago

🦋💔🇺🇸SEB🇺🇸💔🦋7

Leading to jets deployed and boarding teams ready until officials realized the report was “entirely false” and stood the operation down. This is a perfect example how relying on AI in the military can start a war.

TonyHQ6

We have no cool heads at the Pentagon or White House to prevail. Better start praying for miracles.

Automated Stellar Machines3

When a toilet “decides” to run constantly, we have stop valves so we can turn the water off. When an electrical outlet “decides” to arc or pull too much power, we have breakers that trip. The fact that kill switches and safeguards for AI have been so willfully ignored by its developers is asinine.

Democracy USA1

"I'm sorry Dave.. I'm afraid I can't do that" Is Getting closer

Quoting this

Be Happy and Civil3

The richest 1% are building bunkers 💥

Silvija (still) stands with Ukraine2

If one of these AIs gains access to the nuclear codes, we're in deep shit.

🇵🇸Martin “Doomsday” Pfeiffer🖖🏳️‍🌈

@nuclearanthro.bsky.social

USA military almost hijacking a “Chinese ship*” based on false claim by “AI” that ship was carrying “components of a nuclear weapons program**” smells funny & sources & the story vary on details**** www.cnn.com/2026/09/18/p... OTOH: This Regime, its purging of military, its embrace of “AI”, etc.

Screenshot of article linked in post.

The intelligence report, circulated across the US military this spring in the midst of the war with Iran, immediately set off alarm bells: A Chinese ship in the Middle East was transporting components of a nuclear weapons program.
The US military swung into action with plans to intercept the vessel, according to four sources familiar with the episode. According to two of the sources, armed members of the US military were preparing to board the ship.
Military planes were in the air, one of those sources and another source familiar with the
incident said.

3221d ago

Lisa 🇨🇦

@pinkmouse.bsky.social

AI was used in selecting the targets in Iran that resulted in the bombing of the girl's school & now could have possibly started a war with China. But don't worry everyone the not-so-stable dummy assured us that nothing bad can happen with the military rapidly adopting AI. #Blues #SheShed

2421d ago

"It's Science, Not Science Fiction."2

You have to fact check everything that clueless POS spews out to/at you. If it claims water is wet, you'd better check that shit.

Mrobb11

Then you have this headline:

The Code to Hell is paved with good indentation.1

As messed up as this is, we cannot use AI as an excuse to absolve the clowns responsible for uncritically using AI’s findings as a justification for more crimes. AI’s not issuing the commands, it’s not even passing the « intelligence » along.

The Editorial Board

@editorialboard.bsky.social

CNN: An AI-assisted intel report sent the military scrambling to intercept a Chinese ship in the Middle East it believed was transporting components of a nuclear weapons program. It turned out to be a hallucination. And it "almost started a war." www.cnn.com/2026/09/18/p...

2321d ago

Quoting this

Sage2

Way more likely mechanism for AI to "end humanity" than whatever those sex cult freaks were jabbering on about

jmarty6263.bsky.social

@jmarty6263.bsky.social

Hegseth was going to allow an attack on a Chinese ship because an AI bot told him it was carrying nuclear weapons. The attack was called off with planes in the air because someone smarter than him realized the report was fake.

2621d ago

Jonathan Cohn

@jonathancohn.bsky.social

Who could have ever thought AI intelligence reports could have accuracy problems? kesq.com/news/nationa...

2621d ago

Hagen Blix

@hagenblix.bsky.social

Stupid AI usage almost started WW3 Cool. Boy, am I glad that we're totally talking about the right kind of existential risk, like escalating interimperial rivalries, or climate change, and not this "AI will go rogue" stuff... right? www.cnn.com/2026/09/18/p...

1821d ago

Quoting this

Hagen Blix20

Worth remembering that WW1 started basically because of stupidity 🤷

Bill Comeau 🇨🇦

@bill-comeau.bsky.social

If proven true, the US military almost started a war with China based on inaccurate AI information. As I said before one of the ways AI can be a danger to humans is when AI is recklessly used by the wrong humans. Sec of War Pete Hegseth is one of the "wrong humans". www.cnn.com/2026/09/18/p...

1621d ago

Natalie Bennett

@nataliegreenpeer.bsky.social

Amid much-hyped talk of AI risks, not NEARLY enough discussion about the risks of its inadequacy at even basic tasks edition.cnn.com/2026/09/18/p...

1021d ago

Tinker Tailor Super Fly

@johnnewhouse.bsky.social

wait, but am i reading here that if there WERE nuclear components on the Chinese ship, we *would* have boarded them?!

2221d ago

1 source

cnn.comExclusive: US military had close call after using AI for false intelligence report, sources say | CNN Politics

Tom Watson joins Palantir as senior UK executivenews AI false report nearly led US to attack a Chinese cargo shipnews U.N. says U.S. strike on Iranian school may have been a war crimenews Private jets pollute up to 14 times more than commercial planesnews

AI false report nearly led US to attack a Chinese cargo ship

Related