China has tested upgraded AI systems during deep-sea trials of its Jiaolong crewed submersible, demonstrating technology that could eventually support military submarine operations.The trials took place in the Philippine Basin in the western Pacific Ocean, where the submersible reached a depth of 7,056.9 meters (23,000 feet).The upgrades include AI-powered seabed analysis and a fault-diagnosis system designed to assist pilots during underwater operations.The technology could also help improve decision-making and emergency responses aboard nuclear-powered submarines.China tests submarine AIChina has demonstrated AI technology during deep-sea trials of its upgraded Jiaolong crewed submersible, highlighting systems that could eventually support decision-making and emergency responses aboard nuclear-powered submarines.According to Chinese media reports, the trials followed a major upgrade that introduced AI-powered seabed analysis and a separate fault-diagnosis system to assist pilots during underwater operations.The latest development highlighted AI’s potential to support complex underwater missions, where limited communication with the outside world makes autonomous analysis and rapid decision-making particularly important.One of Jiaolong’s upgrades is a fault-diagnosis and decision-support system designed to help pilots identify equipment problems and respond to malfunctions during dives. When a fault occurs, the system provides troubleshooting procedures, records of previous failures and relevant technical documents. It covers critical equipment, including propulsion and control systems, helping pilots identify potential problems and determine appropriate responses.The technology is designed to speed up fault identification and help crews address equipment issues before they develop into more serious problems. Such capabilities could be useful aboard nuclear-powered submarines undertaking prolonged underwater missions. Equipment failures in these environments require immediate attention, particularly when crews have limited access to external technical support. AI-based predictive maintenance could also help identify abnormal operating conditions and potential component failures before they cause equipment breakdowns.Submarines face additional challenges because they cannot depend on continuous satellite communications while submerged. AI-assisted systems could help crews process information, assess equipment conditions and make decisions with limited external support. Jiaolong has also received an intelligent exploration system developed using a multimodal large AI model. The technology analyses images captured by the submersible’s cameras to identify marine organisms and seabed materials.Deep-sea explorationResearchers trained the system using more than 800,000 images of deep-sea organisms and underwater materials. It can interpret camera footage in real time and automatically record identified objects alongside their geographical coordinates, observation times and diving depths.The system also supports manual corrections when it misidentifies objects. Researchers can feed these corrections back into the model to improve its recognition capabilities. Users can interact with the software through dialogue to obtain information about underwater observations. It can also reconstruct the submersible’s diving trajectory in three dimensions, analyze marine species abundance and generate reports on seabed ecosystems.Although developed primarily for scientific exploration, similar AI techniques could potentially be adapted for underwater military operations. For instance, specialized models could be trained to analyze acoustic data collected by submarine sonar systems. These systems receive signals from vessels, marine life and surrounding underwater environments, making accurate classification important for identifying potential threats.According to experts, AI could help distinguish between different sources of underwater noise and flag unusual acoustic patterns for human operators. However, Jiaolong’s current exploration system focuses on visual analysis rather than military sonar-based target recognition. The specific depths at which its AI systems were tested were also not disclosed.The latest trials additionally verified the reliability of Jiaolong’s upgraded equipment, including key components replaced with domestically developed alternatives.