x.com
any explanation as to how and why it took so long for a human to question the data?
The delay stemmed from how the report was produced and the broader operational and institutional context in which it circulated. A Special Operations Command Pacific analyst queried a chatbot (unclear if commercial or government-built) about intelligence reporting on the Chinese ship’s manifest. The tool fused open-source intelligence with classified signals intelligence and incorrectly identified the cargo as components of a nuclear weapons programme. The analyst then used AI a second time to format those findings into a standard intelligence report—the type military officials are accustomed to trusting—and disseminated it widely. Once it looked like a conventional, authoritative product, it immediately triggered alarm bells and operational planning during the high-tempo US–Iran war. Armed personnel prepared to board the vessel and military aircraft were already airborne. Only at the final stage—just before the planned operation—did officials dig deeper into the report’s origins and discover the AI involvement and that it was “entirely false.” Several structural factors explain why human scrutiny lagged: Speed and urgency over verification: The military is pushing AI precisely to accelerate decision-making and the “kill chain.” One source summed it up: “AI allows you to get to a bad idea faster.” Wartime pressure amplified this. Decentralised adoption without unified standards: Different units use different tools under varying rules and safety standards. There is no single, consistent process for verifying AI-generated information. Insufficient guidance on human oversight: Sources noted there is “no real guidance for how having a human in the loop will prevent civilian casualties or fratricide,” particularly as AI use in targeting and intelligence analysis expands. Trust in format and generational differences: Packaging the output as a standard report made it appear reliable. Younger analysts, more native to these tools, are described as more likely to trust them uncritically, while even supportive older officials note pressure to produce and disseminate intelligence faster. In short, the combination of an AI-generated product that mimicked trusted formats, the drive for speed in a wartime environment, and the absence of robust, consistent verification protocols allowed the error to propagate until the last operational checks. Hallucinations of this kind are not described as isolated. The Pentagon and Special Operations Command Pacific did not comment on the specific incident.
