Chinese military researchers are developing artificial intelligence-powered torpedoes designed to outsmart even the most advanced submarine countermeasures. With the potential to transform naval warfare, these high-speed weapons not only travel at blistering velocities but also possess unprecedented intelligence—capable of identifying real enemy submarines despite the use of high-tech decoys and acoustic camouflage.
This technological breakthrough, reported by Chinese scientists in April 2025, signals a new phase in China’s growing dominance of undersea warfare. The announcement came via a peer-reviewed paper published in Command Control & Simulation, a Chinese-language defense journal. According to researchers from the PLA Navy Armament Department and China State Shipbuilding Corporation, the new system achieved a 92.2% success rate in differentiating real targets from decoys during classified tests using high-speed torpedoes.
This represents a significant improvement over legacy guidance systems that have traditionally struggled in environments dense with acoustic deception. As global naval powers scramble to modernize their fleets, this Chinese advancement could force the United States and its allies to rethink their undersea defense strategies.
Modern submarine warfare increasingly relies on advanced decoys and electronic countermeasures to mislead incoming torpedoes. Submarines can deploy acoustic holograms, bubble trails, and other illusion-generating mechanisms to simulate evasive maneuvers or create ghost signatures on sonar screens. These decoys are particularly effective against ultra-fast supercavitating torpedoes—missiles that ride a gas bubble to reduce friction and achieve speeds of up to 200 knots.
But such speed often comes at the expense of precision. Supercavitating torpedoes generate significant noise, complicating target identification by overwhelming sonar systems with distorted echoes. This trade-off has posed a longstanding challenge: how to maintain velocity without losing the ability to distinguish real targets from clever distractions.
Chinese researchers, led by engineers Wu Yajun and Liu Liwen, have developed a method that integrates machine learning with principles of underwater physics to overcome this issue. They began by modeling the behavior of decoys through hydrodynamic simulations, focusing on turbulence and bubble collapse—two key factors in acoustic signature formation. With limited real-world combat data available, this synthetic data was essential to training their AI.
The team employed a sophisticated machine learning technique known as a Generative Adversarial Network (GAN). In this setup, one AI system (the generator) simulates increasingly realistic decoy profiles, while the other (the discriminator) learns to detect their flaws. Over multiple training cycles, the system refined its ability to recognize even the most convincing acoustic forgeries.
To process sonar data, the AI uses a neural network architecture inspired by image recognition. Incoming sound signals are filtered and normalized, then transformed into spectral “thumbnails” using a Fourier transform—a mathematical technique for analyzing frequencies. These are processed by convolutional layers designed to detect irregularities in frequency modulation, while pooling operations smooth out distortions caused by bubbles and interference.
The result: detection rates jumped from 61.3% to over 80% when faced with the most advanced decoys. The system reached over 92% accuracy in more controlled test environments.
This advancement coincides with a rapid buildup in China’s undersea capabilities. The People’s Liberation Army Navy (PLAN) has invested heavily in nuclear and diesel-electric submarines, many of which are equipped with cutting-edge weapons systems.
Among the most notable is the Type 096 nuclear-powered ballistic missile submarine, which is expected to carry the JL-3 intercontinental-range submarine-launched ballistic missile (SLBM). These submarines represent a potent second-strike capability for China, crucial to its nuclear deterrence strategy.
Meanwhile, attack submarines in the fleet are being outfitted with advanced missiles like the YJ-18 anti-ship cruise missile and the CJ-10 land-attack cruise missile. Combined with stealth enhancements, these platforms are designed to project power throughout the Indo-Pacific region and challenge U.S. naval supremacy.
With the integration of AI-powered torpedoes, Chinese submarines could gain a critical edge in contested waters such as the South China Sea and the Taiwan Strait. Enhanced targeting capabilities would reduce the effectiveness of decoys deployed by U.S. and allied submarines, making it harder for adversaries to evade attack.
Interestingly, the official research report cites Operation Hadal (also known internationally as Operation Leviathan), a 2025 Chinese military-themed blockbuster film, as illustrative of the future of submarine combat. The film showcases intense underwater action, with submarines deploying advanced decoys to deceive enemy torpedoes.
In one sequence, submarines emit bubble trails and acoustic illusions to simulate evasive maneuvers, creating a false sonar narrative. These tactics mirror real-world countermeasures in use today, underscoring the increasing sophistication of naval deception technologies.
The researchers note that the AI system they developed is explicitly designed to counter such tactics, relying on deep learning to detect subtle differences between real and fake acoustic signatures.
China is not the only country pursuing AI-enhanced torpedo technologies. Russia’s VA-111 Shkval and various U.S. programs have focused on supercavitation and smart guidance. However, most of these systems struggle with target discrimination at high speeds, leaving them vulnerable to sophisticated decoys.
China’s reported success in solving this issue places it ahead of the curve. According to the researchers, their approach sets “a technical foundation for field deployment” of autonomous underwater vehicles (AUVs) capable of operating independently in noisy and contested environments.
In combat scenarios, where external communication may be disrupted or impossible, these smart torpedoes must make split-second decisions without human input. The AI system addresses this challenge by operating with a high degree of autonomy, reducing the risk of error and wasted firepower.
Moreover, the enhanced recognition system could be integrated into multiple weapon platforms, including both crewed and uncrewed submarines. As China expands its development of unmanned underwater vehicles (UUVs), such AI-guided torpedoes could form part of a vast underwater network designed for surveillance, deterrence, and precision strikes.
For the United States and its regional allies—Japan, Australia, and India in particular—the emergence of such technologies presents a significant threat. The Indo-Pacific theater is already a flashpoint of naval competition, with contested territories and freedom of navigation operations creating near-constant tension.
If Chinese torpedoes can reliably distinguish real targets from decoys, traditional methods of submarine evasion could be rendered obsolete. This would severely constrain the maneuverability and survivability of U.S. submarines operating in the region, potentially undermining their role as a key deterrent.
Additionally, China’s advances in submarine AI reflect its broader military doctrine, which emphasizes “intelligentized warfare”—the use of AI, big data, and automation to gain an edge on the modern battlefield. In this context, the AI-powered torpedo is not a standalone innovation, but part of a comprehensive strategy to modernize and dominate every domain of warfare.
However, questions remain. While the reported success rate is impressive, it’s unclear how the system would perform in real-world conflict scenarios where environmental variables are more chaotic and adversaries deploy novel tactics.
The United States and its allies are unlikely to stand still. Expect a surge in investments into counter-AI technologies, including decoys designed to mimic not just acoustic signatures, but the behavioral patterns that AI systems associate with real submarines. Electromagnetic countermeasures and AI jamming may also come into play.
In response, navies may also double down on stealth technologies, deploying quieter submarines with reduced acoustic footprints, or developing more agile torpedoes capable of counter-countermeasures. The underwater battlefield is poised to become a complex arena of AI versus AI, with rapid innovation cycles and escalating capabilities.
China’s AI-powered torpedoes mark a significant evolution in submarine warfare. By combining supercavitation with advanced target recognition systems, they potentially eliminate the longstanding trade-off between speed and accuracy. More importantly, they signal a shift toward autonomous, intelligent weapons that can operate independently in hostile environments.
As Beijing continues to modernize its submarine fleet and integrate these smart weapons into both crewed and uncrewed platforms, it is reshaping the strategic calculus of undersea warfare. The U.S. and its allies must now contend with a new generation of undersea threats—ones that think, adapt, and strike with unprecedented precision.
