AI Almost Started World War III This Spring

Briefly. Neural network by mistake mistook the cargo of a Chinese vessel for nuclear weapon components for Iran and passed this information to the U.S. military. They immediately scrambled aircraft and began preparing to intercept the vessel, but at the last moment decided to double-check the AI's report: it turned out to be “completely false”. Reports say that if this had happened, it would have plunged the world into a “major war.” Meanwhile, the U.S. military uses AI at every level, including selecting targets for strikes.

U.S. military intelligence regularly experiences failures caused by “hallucinations” in the tools it uses, while young analysts trust algorithms without sufficient critical scrutiny. According to CNN sources, the agency has faced similar problems before. We examine why overreliance on technology and time pressure are leading to an increase in fatal errors.

According to CNN sources, U.S. military intelligence has repeatedly documented cases of “hallucinations”—distortions and errors produced by automated tools. The sources emphasize that similar incidents have occurred within intelligence agencies before, so the problem is not new.

The human factor is becoming a key source of vulnerability in the system. Analysts, especially younger professionals, have grown so accustomed to technological tools that they tend to trust the results they produce without critical analysis or additional verification.

The situation is significantly exacerbated by constant pressure and demands to complete work under extremely tight deadlines. The drive to do everything faster directly reduces the quality of analysis and substantially increases the risk of errors in decision-making.

Note: the source excerpt provided contains only general information about tool “hallucinations,” the behavior of young analysts, and the impact of time pressure, but does not disclose the specific names of AI systems or details of military operations.