Chinese coastguard has 0.3% chance of using weapons in disputed waters, war game suggests

A Chinese coastguard ship was protecting a group of fishing boats when a group of smaller, faster vessels began circling, sometimes moving closer and other times retreating. This scenario posed a tense situation, as it was unclear whether these vessels were merely curious fishermen or attempting to lure the patrol ship into a trap.

Chinese coastguard researchers utilized a computer war game to simulate this high-pressure situation, and the results were surprising. Under the current rule-based command system, the simulated Chinese coastguard would only resort to the use of lethal weapons 0.3% of the time. This low risk, however, was still deemed too high when dealing with sensitive, disputed waters.

To further reduce this risk to zero, the researchers developed an artificial intelligence (AI) framework capable of discerning vessel intentions with greater accuracy than the previous system. This AI framework continuously analyzed the likelihood of various intentions, from routine fishing to coordinated harassment, and adjusted its response accordingly.

In one simulation, the AI-powered system initially assigned a 72% probability that the nearby ship was merely passing by harmlessly. However, when three ships quickly formed a wedge and began closing in from different directions, the system's assessment flipped, raising the probability of "successive probing" to 68%. The AI then ordered the deployment of a water cannon to deter further advances, a measured non-lethal response that adhered to the rules of engagement.

Across multiple simulations, the AI-powered system never triggered a weapons-use violation, effectively bringing the risk down to zero. Additionally, the overall rate of rule breaches – such as getting too close to another vessel – decreased from 1.2% to 0.8%. The AI system outperformed the older rule-based system, which correctly identified a vessel's intentions only 68.5% of the time, whereas the AI achieved a 89.6% accuracy rate.

The researchers attribute this improved performance to the AI's ability to merge radar tracks and other sensor data with a "causal model" – a digital brain that reasons about a vessel's intent rather than merely reacting to its movements. One test involved simulated fishing boats pretending to retreat to distract the coastguard ship, only to have vessels from another direction suddenly accelerate in what appeared to be a collision attempt.

The AI recognized the deception and maintained its stance. The lead author of the study, Sun Shengzhi, a professor at the China Coast Guard Academy, noted that the AI framework creates more rational and flexible opponents, effectively exposing weaknesses in current plans and enhancing decision-support tools for human commanders.

The war game focused on a "foreign interference during fishing protection" scenario, a metaphorical reference to the grey-zone confrontations that frequently erupt in contested waters such as the South China Sea. The central challenge was determining the true intentions of small, maneuverable boats, as hostile probing can often blend with routine navigation or fishing activities.

The AI framework not only lowers the risk of accidental gunfire but also makes war games more realistic by incorporating intention recognition, constrained decision-making, and adaptive opponents into a single system. The researchers emphasized that while AI decisions remain aligned with higher-level strategic intentions and ethical norms, further research is needed to ensure that humans are not entirely removed from the decision-making process.