AI hallucinated a nuclear cargo on a Chinese ship

AI hallucinated a nuclear cargo on a Chinese ship — US nearly intercepted it

CNN reported that in the spring of 2026, during the US war with Iran, US military officials stopped at the last minute preparations to intercept a Chinese ship whose cargo was mistakenly "identified" by a chatbot that mixed open sources with secret signals intelligence.

Compiled by Sergey KostenchukPublished 2026-09-18Updated 2026-09-19

2026-09-18 Safety

🛡 AI hallucinated a nuclear cargo on a Chinese ship — US nearly intercepted it

In the spring of 2026, during the US war with Iran, an analyst from the Pacific Special Operations Command ran a ship's manifest through a chatbot: the model mixed open data with secret SIGINT and mistakenly "identified" the cargo. The same AI formatted the conclusion into an intelligence report; a boarding team and aircraft were put on alert — the operation was stopped at the last minute.

🌍 This is a retrieval system failure: without source provenance labeling, the model conflates open data with secret data. There is no unified verification of AI reports in the Pentagon.

👤 For readers — a reason to double-check bot answers against primary data, especially when the model combines sources unavailable to you.

Source 1: https://edition.cnn.com/2026/09/18/politics/us-military-ai-false-intelligence-china-ship Source 2: https://arstechnica.com/ai/2026/09/report-us-almost-boarded-chinese-ship-over-hallucinated-ai-arms-report/

Sources