In the spring of 2026, a special forces analyst had intelligence synthesized by an AI. The chatbot invented a nuclear cargo; planes had already taken off.
The episode, reported by CNN and resumed on September 18, 2026 by TechCrunch, dates back to the spring of 2026, during the conflict with Iran. A Special Operations Command analyst asked a conversational assistant to merge open source data with classified electromagnetic intelligence. The model concluded that a Chinese-flagged vessel was carrying a nuclear cargo — a completely manufactured conclusion.
The second error was formatting: the analyst again relied on AI to cast these erroneous conclusions into the format of a formal report. The document then went up the chain of command with the appearance of a validated intelligence product. The resulting armed operation was cancelled only at the last moment, while aircraft were already in flight.
Jake Steckler, a Govai researcher and former U.S. Army officer, points out what distinguishes this case from the other hallucination incidents documented so far: decisions likely to result in the use of force have lethal consequences, and no verification device has stopped the report until it circulates.
Read the full article on the original source
Go to the source →