An artificial intelligence chatbot generated a completely false intelligence report that nearly triggered a U.S. military operation against a Chinese vessel in the Middle East, according to an exclusive investigation by CNN based on four sources familiar with the incident.
The incident occurred in the spring of this year, amid the war between United States and Iran. The report, prepared by an analyst from the Pacific Command of Special Operations (SOCOM Pacific), based in Hawaii, stated that a Chinese ship was navigating the region carrying components for a nuclear weapons program.
The document circulated throughout the chain of command and triggered an immediate response: according to two sources, armed military personnel were preparing to board the ship.
Military planes were already in the air when, at the last moment, officials who reviewed the report more thoroughly discovered the error.
The analyst had consulted a chatbot for information about the ship's manifest. The tool merged open-source intelligence with classified signal intelligence and reached an incorrect conclusion.
Next, the analyst used AI for a second time to package the findings into a standard intelligence report—the type of document that military officials rely on—and disseminated it with little further verification.
One source described the document as "completely false" and warned that it "almost starts a war." CNN could not ascertain the actual cargo of the ship or whether the chatbot used was a commercial or government product.
A former senior official stated that "the internal tools are mostly copies of commercial products with a facelift."
Any military operation against a Chinese vessel could have had the potential to escalate into an armed conflict between the two powers, at a time when the strategic rivalry between Washington and Beijing is reaching historic levels. The growing Chinese espionage in the region had already raised alarms among U.S. officials.
The episode takes place in the context of a rapid adoption of AI in the military sphere. In January, Secretary of Defense Pete Hegseth launched the Pentagon's "Artificial Intelligence Acceleration Strategy," aiming to integrate these tools into all facets of operations, from intelligence analysis and target selection to logistics and budgeting.
"We will unleash experimentation, eliminate bureaucratic barriers, and demonstrate the necessary execution approach to lead in military AI," Hegseth declared while announcing the strategy.
A source familiar with current military policies warned: "The use of AI in target selection is definitely on the rise, and there is no real guidance on how human oversight will prevent civilian casualties or friendly fire."
Another summarized the risk with a powerful phrase: “AI allows you to reach a bad idea faster.”
However, the effort is decentralized: different units use different tools under varying orders and security standards, without unified norms to verify the information generated.
The autonomous weapons with AI control deployed by the U.S. had already sparked debate about the limits of human oversight in combat operations.
Sources warn that "hallucinations"—factual errors generated by AI—have not been an isolated incident in the intelligence community since these tools began to proliferate. Several sources indicate that younger analysts tend to rely on them uncritically.
Related videos:
Filed under: