Exclusive: AI-Generated Intelligence Nearly Triggered US Operation Against Chinese Ship

This spring, amid the wider conflict tied to the war with Iran, an intelligence product circulating across the U.S. military asserted that a Chinese vessel in the Middle East was carrying components for a nuclear weapons program. The report prompted immediate, urgent plans to interdict the ship: personnel were being prepared to board, and aircraft were launched, according to multiple sources familiar with the episode.

Only shortly before the planned operation did a deeper review reveal a critical problem: the intelligence had been produced with the aid of artificial intelligence. Investigators found that a chatbot consulted by a special-operations analyst had misidentified the cargo on the ship's manifest. CNN was not able to learn what the misidentified cargo was.

How the Error Unfolded

According to sources, the analyst first used a chatbot to synthesize reporting tied to a manifest that originated with U.S. Special Operations Command Pacific. The tool combined open-source material with classified signals intelligence held by government systems and drew an incorrect conclusion about what the vessel carried. The analyst then used AI a second time to format those findings into a conventional intelligence report and disseminated it through normal channels.

'The report was entirely false,' one source said. 'It almost started a war.' The source warned that any U.S. operation against a Chinese vessel risked triggering a direct and potentially escalatory confrontation between the United States and China.

Why This Matters

Defense and intelligence organizations across the government are rapidly adopting AI to process vast volumes of data, accelerate decision-making, assist in targeting, and streamline logistics. Officials argue this pace is necessary to avoid falling behind potential adversaries. In January, Defense Secretary Pete Hegseth announced an "Artificial Intelligence Acceleration Strategy" to expand AI experimentation and access across the Department of Defense.

Secretary of War Pete Hegseth testifies during a Senate Appropriations Committee hearing in the Dirksen Senate Office Building on Capitol Hill on July 21, 2026 in Washington, DC. - Anna Moneymaker/Getty Images

But the episode highlights serious operational risks when AI tools are deployed without consistent verification standards, unified safety protocols, or clear human-in-the-loop safeguards. Multiple U.S. officials described the current effort as fragmented: different units use different tools under varying authorities and safety checks, and there is no single, consistent process for validating AI-generated outputs.

Sources also warned of a growing cultural pressure to move faster. Junior analysts, comfortable with these tools, can be more likely to accept outputs without sufficient skepticism. The type of error often called an AI 'hallucination' — fabricated or misleading content generated by the model — has surfaced in other government uses of these systems as they proliferate.

Lessons and Next Steps

Officials and observers say the incident underscores the need for urgent reforms before AI is further embedded in targeting and operational decision chains. Potential measures include standardized verification procedures, robust audit trails, mandatory human review for high-risk conclusions, stricter access controls for tools that combine classified and open-source data, and improved training for analysts on AI limitations.

US Special Operations Command Pacific and the Pentagon did not respond to requests for comment on the episode.

The close call serves as a warning: while AI can accelerate intelligence workflows and military decision-making, it can also amplify mistakes with potentially grave consequences if adequate safeguards are not in place.