Condoralex 21 September 2026 3 min read 33 Download PDF

The recent defense summit featuring a meeting in Singapore between U.S. Secretary of Defense Lloyd J. Austin III and Chinese Defense Minister Admiral Dong Jun takes place in a scenario of growing technological and geopolitical tensions. Just as the military leaders of the two superpowers seek dialogue channels, the rapid and uncontrolled rush to adopt artificial intelligence in military commands has shown its most dangerous face.

While Chinese Defense Minister Dong Jun called in Beijing for greater international cooperation and shared rules to curb technology risks, an episode of extreme tension revealed exclusively by CNN demonstrated how an algorithm could push the superpowers to the brink of war.

The Averted Crisis: When AI "Hallucination" Nearly Triggered Conflict

This year, amid the crisis related to the conflict with Iran, an intelligence report produced within U.S. armed forces triggered a red alert across all commands: a Chinese ship in the Middle East was allegedly transporting components linked to a nuclear weapons program.

According to the investigation by CNN journalists Katie Bo Lillis and Zachary Cohen, the episode triggered an immediate response:

• The chain of error: An analyst at Special Operations Command had used a chatbot to analyze the ship's manifest (combining open-source and classified intelligence data) and then format the standard intelligence report. The system "hallucinated", completely falsifying the cargo data.
• The operation blocked at the last moment: American troops had already mobilized to intercept the cargo, with fighter jets in the air and armed personnel ready to assault.

Only before the final action did officials discover that the report was "completely false", stopping an operation that could have triggered a direct armed conflict with China.

The Acceleration of Military AI and the Lack of Standards

• Fragmented standards: The governmental approach is decentralized; different departments use different tools without unified guidelines on verifying generated data.
• The pressure on young analysts: Many junior analysts, digital natives of these tools, tend to blindly trust chatbots, accelerating intelligence production but exponentially increasing the margin of error. As summarized by a source: «AI allows you to arrive at a bad idea faster».
• The human factor dilemma: Internal experts acknowledge that AI use in targeting is growing rapidly, but real rules are lacking on how human oversight (human-in-the-loop) can effectively prevent civilian casualties or friendly fire incidents.

Beijing's Response and the Global Context