A new report says an algorithm's threat assessment almost triggered a US strike on a Chinese vessel.
According to a report citing CNN sources, the US military came close to attacking a Chinese ship after AI-assisted analysis suggested it was carrying components for nuclear weapons. The claim points to AI-generated threat assessments playing a direct role in escalating military decision-making toward a potential strike. CNN's sources are anonymous, and the report includes no timeframe, no location for the ship, and no explanation of how the situation was ultimately resolved.
Militaries have leaned harder on AI for surveillance and target identification, and this claim - if accurate - would be one of the first public examples of that automation nearly forcing a confrontation between two nuclear powers. It echoes an old Cold War lesson: automated warning systems have triggered false alarms before, and human judgment has historically been what stopped them from becoming shooting wars.
But an extraordinary claim like this needs more than unnamed sources and a missing timeline. Until someone attaches a date, a location, and an on-record confirmation to it, this reads as a report about a report, not a confirmed near-miss.