A false intelligence report generated with the help of Artificial Intelligence nearly prompted the US military to launch an operation against a Chinese vessel in the Middle East, triggering a crisis that could have escalated into direct confrontation between Washington and Beijing, according to a CNN investigation.

The report, which circulated among the US armed forces in the spring amid the war with Iran, set off alarm bells. It claimed that a Chinese vessel in the Middle East was carrying components linked to a nuclear weapons development program. US military agencies responded immediately, drawing up plans to intercept the vessel. According to sources who spoke to CNN, armed teams were even preparing to board the ship, while US military aircraft had already taken off.

However, shortly before the operation was due to begin, officials took a closer look at the report and discovered that it had been prepared with the help of an Artificial Intelligence tool, which had incorrectly identified the vessel’s cargo.

According to CNN, it is not known what the actual cargo was that had been misinterpreted by the AI system. One source, however, described the report as “completely false” but said it had “almost started a war”, as any US operation against a Chinese vessel could have triggered a serious military escalation.

How AI created the crisis

The incident highlighted the risks posed by the rapid integration of Artificial Intelligence into US military and intelligence agencies. In this case, an analyst used a chatbot to process information about the vessel’s cargo obtained from US Special Operations Command Pacific, based in Hawaii.

The system combined open-source data with classified signals intelligence held in US government databases and reached an incorrect conclusion about the nature of the cargo. The analyst then used Artificial Intelligence again to generate a formal intelligence report, which was circulated to military officials as verified analysis.

It was not made clear whether the chatbot used was a commercially available AI tool or a system developed by the US government. “Internal tools are largely just copies of commercial applications, with a more ‘polished’ wrapper,” explained a former senior US official familiar with the AI systems used by military and intelligence analysts.

The Pentagon and US Special Operations Command Pacific did not comment on the incident.

Pressure for rapid adoption of AI in the military and fears of fatal mistakes

The US armed forces are investing increasingly heavily in Artificial Intelligence in an effort to accelerate decision-making on the battlefield, from target selection to the management of military resources, supply chains and budgets.

The US government maintains that the country cannot afford to fall behind in the development of military AI, particularly in the face of adversaries such as China. In January, Defence Secretary Pete Hegseth unveiled the Pentagon’s strategy for accelerating the use of Artificial Intelligence, stressing that the US must strengthen its capabilities in the field. “We will unleash experimentation, eliminate bureaucratic barriers and focus our investments to ensure that we lead in military AI technology,” he said.

However, officials familiar with how the systems operate argue that they are being deployed without a unified safety framework. Different agencies use different tools, with different oversight rules and varying levels of reliability.

The incident involving the Chinese vessel is not, according to CNN sources, an isolated example. So-called AI “hallucinations” — the generation of convincing but inaccurate information — have also occurred in other cases within the US intelligence community.

Particular concern has been raised over the use of AI in targeting operations, where a mistake can have immediate human consequences. “AI in targeting is certainly increasing, and there is no real guidance on how having a human in the loop will prevent civilian casualties or friendly fire,” a source familiar with current military policies told the US network.

At the same time, some former intelligence officials are concerned that pressure to produce analysis more quickly could lead to more errors, particularly when younger analysts place excessive trust in AI tools. As one of the sources who spoke to CNN put it: “AI allows you to get to a bad idea faster.”