In the spring of 2026, U.S. military forces were poised to intercept a Chinese vessel in the Middle East based on a false intelligence report generated by artificial intelligence, as reported by CNN on September 18.
The report claimed that the ship was transporting components linked to a nuclear weapons program. According to four sources cited by CNN, military personnel were ready to board the vessel, with aircraft already in the air.
However, prior to the operation, more experienced analysts re-evaluated the initial data and identified the mistake. It remains unclear what the ship was actually carrying, with one source describing the report as "completely false."
The U.S. Special Operations Command and the Pentagon did not respond to media inquiries regarding the incident.
How the Error Occurred
An analyst within the special operations unit uploaded intelligence data about the ship's manifest into an AI chatbot, which had been sourced from the U.S. Pacific Command. It is unclear whether the analyst used a commercial model or a government internal system.
The neural network mistakenly matched publicly available information with classified electronic intelligence data, leading to an incorrect cargo identification. The analyst then used AI again to format conclusions into a standard intelligence report and submitted it to the military.
According to CNN’s sources, the use of such tools within the military and intelligence remains decentralized, with various entities employing their own systems and security protocols. There is no unified standard for verifying generated information.
One source highlighted that similar "hallucinations" have previously occurred with AI usage by American intelligence agencies:
“Artificial intelligence can quickly lead to poor ideas.”
The military is increasingly utilizing neural networks for analyzing intelligence data and selecting targets. One source from CNN stated that the application of AI in these tasks is expanding, though clear guidelines regarding human oversight to prevent erroneous strikes are still lacking.
It is worth noting that in June, Bloomberg reporters discovered that the Pentagon had secretly approved an updated doctrine for utilizing AI in target selection on the battlefield.