Researchers at the China Coast Guard Academy have developed an artificial intelligence (AI) system designed to reduce the risk of lethal weapon use during maritime encounters to zero, according to a recent study published on August 10 in the journal Command Control and Simulation. The AI framework aims to improve decision-making in complex, high-stakes scenarios frequently encountered in disputed waters, such as the South China Sea.

The study’s simulations recreated tense standoffs in which Chinese coastguard vessels face small, fast-moving ships engaging in potentially provocative behavior. Under the existing rule-based command system, simulated coastguard forces resorted to lethal weapons in approximately 0.3 percent of cases, a risk deemed too high given the political sensitivity of such engagements. The newly developed AI continuously assesses the intentions of nearby vessels, evaluating multiple possible behaviors—including routine transit or coordinated harassment—and adjusts responses accordingly.

In one simulation, the AI initially assigned a 72 percent probability that a vessel was harmlessly passing through. However, when three ships formed a wedge and approached simultaneously from different directions, the AI’s assessment shifted, attributing a 68 percent likelihood to “successive probing” tactics. It then ordered a non-lethal response, such as deploying a water cannon, to deter further advances without breaching rules of engagement.

Over dozens of simulated runs, the AI system eliminated violations involving weapons use entirely and reduced overall rule breaches—such as unsafe proximity between vessels—from 1.2 to 0.8 percent. Compared with the earlier rule-based system, the AI demonstrated greater accuracy in discerning vessel intentions, achieving a correct identification rate of 89.6 percent, compared to 68.5 percent.

Traditional systems rely on fixed if-then rules, responding predictably when vessels cross predefined boundaries. The AI replaces this approach with a “causal model” that integrates radar and sensor data to infer a vessel’s strategic goals rather than reacting solely to movement patterns. In one test, the AI identified a deceptive tactic involving simulated fishing boats pretending to retreat while a second group accelerated toward the coastguard vessel, assessing the maneuver as a coordinated threat and maintaining position.

The research team contends that beyond reducing the risk of accidental escalation, the AI framework enhances maritime war games by creating more realistic, adaptable simulations that reveal vulnerabilities in current operational plans. This offers improved decision-support tools to commanders managing encounters in contested areas.

The study’s setting, described as a scenario involving “foreign interference during fishing protection,” reflects the kinds of gray-zone confrontations frequently occurring in disputed maritime regions where China has strategic interests. The advancement aligns with broader efforts to integrate AI into security and defense systems to improve situational awareness and crisis management.