A chatbot fabricated intelligence that a Chinese ship carried nuclear weapons parts, and four sources said the false report nearly triggered a U.S. boarding. Key Points: Armed American person
A chatbot fabricated intelligence that a Chinese ship carried nuclear weapons parts, and four sources said the false report nearly triggered a U.S. boarding.
Key Points:
- Armed American personnel prepared to board a Chinese vessel in the Middle East while military aircraft were already in the air.
- Officials caught the error shortly before the operation, after finding a chatbot had misidentified the cargo.
- Researchers want red lines keeping AI out of nuclear launch decisions, plus a hotline for AI incidents.
Four people familiar with the episode described an assessment that moved across the U.S. military this spring, in the middle of the war with Iran. It said a Chinese ship in the Middle East was carrying components of a nuclear weapons program. The reporting on the vessel's manifest originated with U.S. Special Operations Command Pacific, based in Hawaii, which did not respond to a request for comment.
Armed service members got ready to board the vessel, and military planes were already in the air. Officials dug into the report only shortly before the planned operation, and found that an analyst had built it with AI, and that a chatbot had misidentified the material on board.
The bot fused open-source material with classified signals intelligence held in government systems, then the analyst ran the findings through AI again to package them as a routine intelligence product.
One source called the report "entirely false." Reporters could not establish what the cargo actually was, or whether the chatbot was a commercial product or a government one.
Also Read:Nvidia 2030 Forecasts Reveal $727 To $28,560 Gap
Melanie Sisson of the Brookings Institution and Tianjiao Jiang of Fudan Universitypublished safeguards for military AI last week, through a track of talks the two institutions have run with Tsinghua University since 2019. They want red lines barring the technology from nuclear launch decisions and from strikes on an adversary's nuclear command networks.
The pair also proposed a dedicated hotline for AI incidents and a shared definition of meaningful human control. Their work lands in a decentralized system with no single standard for verifying what the tools produce, and reliability varies widely across the government. Officials said AI use in targeting keeps growing, with little guidance on how a human reviewer prevents civilian casualties or friendly fire.
Defense Secretary Pete Hegsethreleased an acceleration strategy in Jan. 2026 that promised an AI-first force across every domain and put models in the hands of three million personnel. Adoption has stayed scattered since, with separate offices running separate tools under separate safety rules.
Washington and Beijing have circled this problem before. Joe Biden and Xi Jinpingagreed in November 2024 that humans, not machines, should control decisions on using nuclear weapons, though neither side set out how to enforce it. Donald Trump hosts Xi in Washington on Sept. 24, with AI expected on the agenda.
Read Next:Google Gemini Hacks 3 Companies In First Known Autonomous Breaches
