The US military came within hours of intercepting a Chinese vessel in the Middle East this spring after an intelligence report turned out to be built on an AI hallucination, CNN reported. A chatbot incorrectly identified material the ship was supposedly carrying as components of a nuclear weapons program. Four sources familiar with the episode said the report was circulated across the US military during the war with Iran and immediately triggered alarm. Plans were drawn up to intercept the vessel, and two sources said armed service members were preparing to board it while military aircraft were already in the air.
Only just before the operation did officials dig deeper and discover a special operations command analyst had generated the report with AI assistance. The analyst queried a chatbot about intelligence on the ship manifest from US Special Operations Command Pacific, and the bot fused open source intelligence with secret signals intelligence before reaching its conclusion. The analyst then used AI again to format the findings into a standard intelligence report and disseminated it. One source described the report as entirely false, and said it almost started a war.
The episode lands as the Pentagon pushes AI into nearly every facet of military work, from analysing raw intelligence and selecting strike targets to budgeting and supply chains. Defense Secretary Pete Hegseth released an Artificial Intelligence Acceleration Strategy in January aimed at getting leading models into the hands of three million personnel. But the rollout is decentralised, with different agencies using different tools under different safety standards and no single verification standard. Sources said AI use in targeting is ramping up with little guidance on how human oversight will prevent civilian casualties or fratricide. The close call is a reminder for founders building agentic systems that confident, well formatted output is not the same as correct output.