WASHINGTON, United States — A false AI intelligence report led the United States military to prepare to intercept a Chinese vessel in the Middle East in spring 2026, according to four people familiar with the episode.
The report wrongly said the ship was carrying components for a nuclear weapons programme during the war with Iran.
Two of the people said the military had assembled armed personnel to board the vessel. Two people also said American military aircraft had already taken flight.
Officials uncovered the error only shortly before the planned operation was due to begin.
Their closer examination found that an analyst at a special operations command had produced the report with help from artificial intelligence, and that a chatbot had incorrectly identified material listed as cargo.
The accounts did not disclose what the vessel was actually carrying.
One person familiar with the incident described the report as “entirely false” and said it “almost started a war.”
The person said American action against a Chinese ship could have escalated into armed conflict between the two countries.
No Uniform Verification Standard
The adoption of artificial intelligence has not followed a single process across the United States military and intelligence community, according to several officials familiar with the effort.
Individual organisations use different systems, operating instructions, and safeguards, the officials said.
There is also no government-wide standard for verifying the information those systems produce, according to the officials.
Performance and reliability vary among the tools deployed by separate military and intelligence organisations.
How the Report Was Produced
The underlying information concerned the vessel’s manifest and originated with United States Special Operations Command Pacific, which is based in Hawaii.
An analyst submitted some of that reporting to a chatbot, according to people familiar with the process.
The accounts did not identify the system or establish whether it came from a commercial provider or the United States government.
“The internal tools are mostly just copies of the commercial stuff wearing lipstick,” a former senior United States official familiar with the artificial-intelligence systems used by military and intelligence analysts said.
The chatbot drew on publicly available material and classified signals intelligence held by the government, according to one person familiar with the episode.
From those sources, it produced the incorrect finding about nuclear-related cargo.
Artificial intelligence was then used to convert the findings into the military’s conventional intelligence-report format.
The analyst distributed the resulting document within the military.
Military Adoption Accelerates
Military and intelligence officials are introducing artificial intelligence into a broad range of work.
The systems are being used to process large quantities of collected intelligence, assist in identifying possible strike targets, and perform administrative tasks involving budgets, logistics, and supply chains.
Officials supporting the expansion say the technology can reduce the time needed to decide which targets to strike or where to position military assets.
They also say the United States could lose an operational advantage if China or another potential adversary adopts the technology more quickly.
Defence Secretary Pete Hegseth, who testified before a Senate Appropriations Committee hearing in Washington on Tuesday, July 21, 2026, released the department’s “Artificial Intelligence Acceleration Strategy” in January.
“We will unleash experimentation, eliminate bureaucratic barriers, focus our investments and demonstrate the execution approach needed to ensure we lead in military AI,” Hegseth said when he announced the strategy.
A memorandum accompanying the plan directed the department to expand access through several programmes.
Its stated objective was “democratizing AI experimentation and transformation across the Department by putting America’s world-leading AI models directly in the hands of our three million civilian and military personnel, at all classification levels.”
Targeting and Human Review
The military’s use of artificial intelligence to help select targets is increasing, according to people familiar with current policies.
One of them said existing guidance did not explain how human review should be used to prevent the killing of civilians or American and allied personnel when automated systems contribute to targeting decisions.
“AI in targeting is definitely something that is ramping up and there is no real guidance for how having a human in the loop will prevent civilian casualties or fratricide,” the person said.
One source said the false output in the Chinese ship episode was not an isolated occurrence after artificial-intelligence tools began spreading through the intelligence community.
Several people familiar with their use said the technology had increased pressure on analysts to prepare and circulate intelligence more quickly.
Younger analysts, who tend to have greater familiarity with the tools, may also be more inclined to accept their results without sufficient scrutiny, the people said.
“AI allows you to get to a bad idea faster,” one of the sources said.
