US military had close call after using AI for false intelligence report

https://edition.cnn.com/2026/09/18/politics/us-military-ai-false-intelligence-china-ship

They almost attacked a Chinese ship believing it was transporting nuclear weapons component (/via OSINTTechnical@X)

The future will be much more embarassing for humanity than Skynet

@buherator AI hallucination causing WWIII is on my 2026 bingo card.

@buherator i remember reading during the second iraq war, operations led by the british had far fewer friendly fire casualties because they had not just a system for putting a case for a target together, but also researchers trying to rule out the validity of a target (that's a friendly, that's a school, etc)

the americans didn't have that

...then they automated it

and i guess, *now* the stakes are high enough to give them a target offramp, huh

@buherator (but as long as it's not what they do every day, it's only a matter of time)

*edit* this being fedi i'm editing to point out the also obvious context that ai has a tendency to hallucinate "facts" and conclusions that aren't supported by anything

@frobozz My prediction is that this will be covered by some bullshit about "adjustment" then a few million will be spent on additional guArDrAILs then the next hallucination will cause actual damage. So yeah, not a bad bet.