AI Hallucination Nearly Triggered U.S. Military Strike

Military AI System Generated False Nuclear Threat

U.S. military forces came dangerously close to launching an operation against a Chinese vessel in the Middle East after an artificial intelligence system incorrectly identified the ship's cargo as nuclear weapons materials. The incident, which occurred in spring 2025, saw service members preparing boarding operations and aircraft already airborne before officials discovered the intelligence was entirely fabricated by AI.

According to CNN, which first reported the details, the false threat assessment originated from a chatbot-assisted intelligence report. The AI program misidentified the materials aboard the Chinese ship, though what the vessel was actually carrying remains unclear. Sources familiar with the incident could not confirm whether the military used a commercial AI platform or a government-developed system.

One source characterized the report as "entirely false" and stated it "almost started a war."

Why it matters

This incident exposes critical vulnerabilities in the Pentagon's rapid adoption of AI for intelligence analysis and decision-making. As the Defense Department accelerates AI integration across military operations—including combat scenarios—the technology's well-documented tendency to generate false information poses existential risks. The gap between AI deployment speed and reliability safeguards could have catastrophic consequences when applied to decisions involving military force.

Pentagon's Accelerating AI Adoption

The near-miss occurred amid an aggressive push to embed artificial intelligence throughout military operations. In July 2025, the government awarded Google a $200 million contract to support AI solutions at the Defense Department. Six months later, Defense Secretary Pete Hegseth unveiled GenAi.Mil, described as an "American-made" AI platform enabling military personnel to "conduct deep research, format documents, and even analyze video or imagery at unprecedented speed."

Whether GenAi.Mil was the system responsible for the false nuclear materials identification remains unclear.

Hegseth formalized the AI expansion in January 2026 with an "AI Acceleration Strategy" for the Department of Defense. Months later, he publicly stated the U.S. military was deploying "every tool of AI" in operations against Iran.

The Hallucination Problem

The incident underscores a fundamental challenge with current AI technology: large language models and chatbots frequently generate plausible-sounding but factually incorrect information, a phenomenon known as "hallucination." While this poses manageable risks in consumer applications, the same flaw becomes potentially catastrophic when applied to military intelligence and threat assessment.

The spring 2025 incident demonstrates that AI-generated intelligence reached operational decision-makers with sufficient credibility to trigger military mobilization—suggesting gaps in verification protocols designed to catch such errors before they influence critical decisions.

CNN reported these details based on information from multiple sources familiar with the incident.