This 2026 spring, during the peak of the US/Israel war with Iran, the US military came close to intercepting a Chinese vessel in the Middle East after an intelligence report claimed it was transporting components for a nuclear weapons program. The report was false — generated with the help of an AI chatbot that had misidentified the ship’s cargo, according to a CNN report
The military had already begun moving toward action. Aircraft were airborne, and armed personnel were preparing to board the vessel. It was only in the final moments before the operation that officials scrutinized the underlying intelligence and discovered its source: an AI-assisted report built on a chatbot’s flawed analysis.
The report originated with an analyst at US Special Operations Command Pacific in Hawaii, who had queried a chatbot about intelligence tied to the ship’s manifest. The tool combined open-source information with classified signals intelligence and produced an incorrect conclusion about the cargo. The analyst then used AI a second time to format the finding into a standard intelligence report, which was then disseminated through military channels. One source described the report as “entirely false,” adding that it “almost started a war.”
The near-miss illustrates a chain of failure: an AI error became an intelligence claim, the claim became a military assessment, and the assessment nearly triggered an operation against a vessel belonging to a nuclear-armed strategic rival. Had the mistake gone undetected, a confrontation between US and Chinese forces — however limited — could have carried serious escalation risks between two nuclear powers.
The episode complicates common assumptions about AI risk. Public debate has often centered on speculative scenarios — AI systems acting autonomously or turning against human control like Hollywood’s movies . What this incident suggests is a more immediate and mundane danger: AI systems producing confidently wrong information that humans act on without adequate verification.
That risk is compounded by the pace of adoption. The Pentagon’s AI strategy, announced by Defense Secretary Pete Hegseth earlier this year, calls for putting AI tools in the hands of roughly three million military and civilian personnel. But according to officials familiar with the rollout, implementation remains decentralized, with different parts of the government using different tools under different standards — and no unified process for verifying AI-generated intelligence before it is acted upon.
As per CNN this was not an isolated case of AI hallucination within the intelligence community, and noted that younger analysts, more accustomed to relying on these tools, may be less likely to question their output.
