US Military Almost Boarded a Chinese Ship After AI Hallucinated a Nuclear Threat
An AI-generated intelligence report nearly triggered a US military operation against a Chinese vessel in the Middle East this spring, according to four sources familiar with the incident reported by CNN. The report, produced by a special operations command analyst with AI assistance, falsely claimed the ship was carrying components of a nuclear weapons program. It was entirely fabricated.
Armed US military personnel were preparing to board the vessel. Military planes were already airborne. The operation was only halted when officials dug deeper into the report’s provenance and discovered it had been assembled using an AI chatbot — one that had confabulated the ship’s cargo.
What the System Did
The analyst, working at US Special Operations Command Pacific in Hawaii, queried a chatbot against intelligence reporting on the ship’s manifest. The bot fused open-source intelligence with classified signals intelligence from government holdings and produced a conclusion about nuclear weapons components that had no factual basis.
The analyst then used AI a second time — to format the findings into a standard military intelligence report, the kind that commanders treat as authoritative. The report circulated across the US military before anyone checked its sourcing.
Whether the chatbot was a commercial product or a government tool is not publicly known. A former senior US official described the distinction as largely cosmetic: “The internal tools are mostly just copies of the commercial stuff wearing lipstick.”
Why It Almost Became a War
The stakes in this case were high specifically because the alleged violator was Chinese. Any US military boarding or seizure of a Chinese vessel in a Middle East war zone carried the risk of direct confrontation between the two nations. One source described the report as something that “almost started a war.”
The incident occurred in the context of an already active war with Iran, when military and intelligence workflows were under pressure and the threshold for acting on intelligence reporting was compressed.
The Broader Pattern
Across the US military and intelligence community, officials have been integrating AI into targeting workflows, intelligence fusion, logistics, and supply chain management. The pressure to process large volumes of raw intelligence quickly has made AI tools attractive precisely in contexts where errors are most consequential.
The hallucination in this case was not a subtle one — it produced a false assertion about WMD proliferation on a vessel belonging to a nuclear-armed state. The error survived two AI-mediated steps: generation and formatting. Neither step flagged the absence of primary evidence for the central claim.
The incident is now feeding internal debates about where human review must sit in the targeting chain, and whether AI-formatted intelligence reports should carry any automatic procedural authority.
No policy changes have been publicly announced. The US military has not commented on the episode.