The American military came dangerously close to an armed confrontation with China over the weekend after an artificial intelligence system made a critical error. According to sources familiar with the situation, the military had already begun preparing an operation to intercept a Chinese vessel — armed soldiers were readying to board the ship, and military aircraft were already in the air.
The scare was triggered by an intelligence report drafted by a special operations command analyst using an AI-powered chatbot. Just before the operation was set to launch, it emerged that the program had misidentified the cargo the vessel was carrying. What the ship was actually transporting remains unclear — one source described the report as "completely false."
"This nearly led to war," said a source familiar with the details of the incident.
The analyst had asked the chatbot for intelligence on the ship's cargo, drawing on data from the US Special Operations Command Pacific, headquartered in Hawaii. It remains unclear whether the analyst used a commercially available chatbot or a tool specifically developed for the US government.
According to a former senior US administration official familiar with the AI systems used by military and intelligence analysts, internal government tools often differ little from commercial products. "Internal tools are mostly just copies of commercial products with a new interface slapped on," the official noted.
The chatbot merged open-source information with classified data obtained through signals intelligence available to the government. As a result, the system drew an incorrect conclusion about the nature of the cargo. That wasn't the end of the AI's involvement, either: the analyst then used the chatbot again to format the findings into a standard intelligence report, which was subsequently circulated among military units.
The incident exposed the downside of the rapid rollout of artificial intelligence across the US armed forces. American authorities plan to use AI, among other things, to sift through vast troves of intelligence and to help select targets for military strikes.
The technology is meant to help the military make faster decisions — for instance, which target to strike or where to redeploy military resources. At the same time, US officials have insisted that Washington cannot afford to fall behind potential rivals, including China, in AI development.
Source: novinky.cz