An American intelligence report built with AI tools mistakenly flagged a Chinese ship as carrying nuclear arms program parts, nearly triggering a US military boarding operation in the Middle East. The incident, described in a CNN report, shows how far AI hallucinations can reach.
Ars Technica’s report was prepared by a US Special Operations Command analyst who used a chatbot to analyze intelligence documents about the ship’s cargo. The chatbot combined public information with secret signals intelligence held by the government, and the result was fed into an official document. Military planners were preparing to intercept and board the vessel with air support when officials realized the chatbot had incorrectly identified what the ship was carrying.
Four sources familiar with the episode told CNN that the intelligence was “entirely false.” One source described the AI-powered mistake as “almost starting a war.”
The Cargo Mix-Up
The core problem was simple: the chatbot confused what the ship was moving. The analyst relied on its output to build the report, and the machine’s error made its way into the final document.
The report’s conclusion that the ship was transporting nuclear arms program components was entirely wrong. The military was poised to act on it before the mistake was caught.
Who Was Involved
The report draws on four sources familiar with the episode, all cited by CNN. The analyst worked for US Special Operations Command, and the military response was coordinated with air support ready to assist the boarding operation.
The chatbot itself was not named in the report, nor was the platform it ran on. What is clear is that the tool was asked to analyze intelligence documents and produce an assessment, and that assessment was wrong.
| Party | Role |
|---|---|
| US Special Operations Command analyst | Prepared the report using a chatbot |
| Chatbot | Combined public and secret intelligence, misidentified cargo |
| US military | Preparing to intercept and board the ship with air support |
| Four sources | Familiar with the episode, cited by CNN |
Why Hallucinations Happen
Since “hallucinating” became the Cambridge Dictionary’s word of the year in 2023, there have been numerous examples of professionals being misled by AI tools that fabricate information when their training data does not provide enough context. Non-fiction authors, journalists, academic researchers, judges, doctors, police departments, and corporate call centers have all fallen victim to machines making things up.
Some researchers suggest it may be impossible to stop large language models from hallucinating altogether, no matter how carefully they are prompted. The attempt to prevent LLMs from hallucinating has produced mixed results, with some systems still producing false claims even when instructed not to.
What the Pentagon Is Doing
The US Department of Defense rolled out an “AI acceleration strategy” in January. The plan sought to “make all appropriate data available across federated IT systems for AI exploitation, including mission systems across every service and component.”
That strategy is now being tested against the reality of AI systems producing false intelligence. The CNN report describes the incident; it says nothing about Pentagon comments.
The Consequence of a Mistake
The incident raises a question that is uncomfortable for anyone who depends on machine-generated intelligence: how do you know what is real? The analyst used a tool designed to help, and the tool produced something that could have set off a military action.
The Pentagon’s January strategy aimed to make more data available to AI systems. This incident suggests the strategy also needs safeguards that catch errors before they reach decision-makers.
The CNN report does not say whether the analyst knew the report was wrong, or whether the chatbot’s output was presented as confirmed fact. What is known is that the report was submitted with the machine’s error intact, and that error put a ship in the crosshairs.
This is one of the more potent and consequential instances of a hallucinating AI ruining the reliability of a professional report.
The lesson from this incident is that AI systems need to be tested and monitored constantly, not just trusted to perform. A machine’s hallucination can travel fast, and the consequences of believing it can be severe.
Source material: “AI hallucination of Chinese nuclear components almost led to US military attack,” Ars Technica.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.