The U.S. military reportedly came close to intercepting a Chinese vessel after an artificial intelligence tool wrongly concluded that the ship was carrying nuclear weapons materials.
A CNN report on Friday claims that military personnel got as far as preparing to board the ship while aircraft scrambled to the site before discovering the intelligence was faulty.
The incident occurred earlier in the spring, in the first months of the U.S.-Iran conflict. According to the report, a special ops analyst used an AI chatbot to gather information about the ship's cargo, which then led to a full-on intelligence report that misidentified what the ship was carrying.
The planned operation was called off once the error was discovered, but one source said the episode "almost started a war."
The near-miss comes as the Pentagon is expanding its use of generative AI. Separate reports from earlier this month claim that more than 1.7 million personnel have used GenAI.mil since the Department of War, headed up by Pete Hesgeth, launched it nine months ago.
The internal platform gives military and civilian employees access to AI models for research, document analysis, policy work and logistics.
ChatGPT Mil and Grok for Government were also recently added to the platform, with both approved to handle Controlled Unclassified Information. The CNN report, however, did not say whether or not the chatbot was involved in the Chinese ship incident.
Hours after CNN published the report, President Donald Trump announced on Truth Social that CNN, MS NOW and Politico would be barred from the White House "effective immediately."
He went on to accuse the outlets of publishing "fiction and lies," though no direct connection between the ban and this particular story was established.