US military came close to attacking a Chinese ship due to an artificial intelligence error — CNN

US military came close to attacking a Chinese ship due to an AI error — CNN

CNN, citing sources, reports that the US military came close to launching an operation to intercept a Chinese ship this spring due to an error in artificial intelligence.

According to the TV channel, a US military intelligence analyst used a chatbot to analyze data regarding a Chinese ship's cargo manifest. Combining open-source information with classified intelligence, the AI drew an incorrect conclusion that the ship was carrying parts related to Iran's nuclear weapons program.

Upon receiving the report, the US military immediately began developing an action plan. Military personnel were prepared to board the ship, and military aircraft were scrambled.

However, shortly before the start of the operation, officials decided to review the report in detail. It was then revealed that the document had been prepared using artificial intelligence and that the chatbot had incorrectly identified the cargo being transported.

One of CNN's sources described the report as "completely false and almost starting a war." The TV channel noted that any US military operation against a Chinese ship could have led to an armed conflict between the two countries.

According to CNN, US officials are seeking to expand the use of artificial intelligence in almost all areas of military and intelligence systems, including processing "raw data" and identifying strike targets.

In this case, the military analyst first had the chatbot analyze intelligence data regarding the ship's manifest. Then, using artificial intelligence again, they prepared a report and presented it to senior leadership.

The US Department of Defense has been actively expanding the use of artificial intelligence in the military sphere in recent years. Last year, the Pentagon signed a $200 million contract with Elon Musk's xAI company to accelerate the integration of artificial intelligence into the department's operations.

The US Department of Defense has also signed contracts with OpenAI and Anthropic.

Earlier, The Wall Street Journal wrote that the Donald Trump administration used Anthropic's artificial intelligence technology to plan a major airstrike on Iranian territory.

Later, a conflict arose between the Pentagon and Anthropic. The company's head, Dario Amodei, demanded guarantees that artificial intelligence technology would not be used for surveillance or autonomous weapons without human supervision.

At the end of August, the Pentagon's head of research and development, Chief Technology Officer Emil Michael, stated that the Department of Defense plans to completely stop using Anthropic's artificial intelligence in its systems by the end of September.