In the Terminator films, humanity launches an all-out assault in 2029 on Skynet’s central node deep within Cheyenne Mountain – only to realise that the artificial intelligence has no single point of failure.

Skynet is not a monolithic system, but a distributed, self-healing network, capable of surviving total infrastructure collapse and resurrecting itself almost instantly.

Today, China appears to be building a real-world analogue – a highly resilient, AI-driven combat kill web composed of thousands of uncrewed aerial, surface and underwater platforms.

This network is designed to persist through destruction, bearing an unsettling resemblance to Skynet’s structure and survivability.

At a defence technology conference in Tianjin this month, engineers from Tianjin 712 Communications & Broadcasting Co – a key supplier of military communication systems – revealed breakthroughs in elastic mesh networking for unmanned combat clusters.

Field test results suggest the new battlefield network is so robust and adaptive that it could continue even after suffering massive damage – the first time such data has been released to the public.

control, adaptive routing, and AI-driven topology optimisation, the smart network could reconfigure itself in under 0.1 seconds when key nodes were destroyed, according to a conference paper titled “Unmanned Cluster Elastic Networking Communication Technology and Its Application in Low-Altitude Operations at Sea”.

When facing severe jamming in electronic warfare, this decentralised mesh network allows swarms of drones, uncrewed boats and undersea vehicles to maintain over 80 per cent of their communication capacity.

“This technology differs from traditional fixed network architectures in that it emphasises ‘elasticity’ – the ability of the network to deform under external pressure, like a rubber band, without breaking, and to return to its original state once the stress is removed,” wrote the team led by defence communication expert Hu Haiyang.

At present, PLA’s kill web remains fundamentally different from Skynet. It does not seek self-preservation or global domination.

But scientists are improving its intelligence and strength.

“Future research will focus on exploring the potential applications of quantum communication technology in unmanned swarms,” Hu’s team wrote.

They were also developing smarter autonomous cooperative communication algorithms and studying integrated designs that unified communication, sensing and computing to “broaden the application of the technology in military and civilian sectors”, they added.

The PLA kill web employs dynamic, self-healing topologies using game theory and reinforcement learning algorithms to reassign roles and relay paths within “milliseconds”.

Cognitive radio and cross-layer protocol optimisation enables the network to sense environmental changes such as interference or node loss and automatically switch frequencies, modulation schemes or routing strategies, according to Hu and his colleagues.

Deep learning models also help detect jamming signals with unprecedented speed and deploy countermeasures such as frequency hopping, beamforming or transmission timing randomisation.

These capabilities enabled “saturation attacks” from multiple domains – air, sea and underwater – where drone swarms could autonomously coordinate low-altitude penetration, electronic warfare, reconnaissance and precision strikes, even if command links were severed or key platforms destroyed, Hu’s team said.

For example, swarms of cheap drones or loitering munitions could launch coordinated assaults on enemy warships or coastal defences, maintaining cohesion via ad hoc mesh links when parts of the swarm were shot down.

The technology also makes wide-area collaborative reconnaissance a reality, with clusters of high-altitude unmanned aerial vehicles (UAVs), surface drones and underwater sensors fusing radar, optical, acoustic and electronic data into a unified battlefield picture, resilient to partial node failure.

The Chinese military’s interceptor drones, lasers and railguns could also use the network to form a layered defence grid, sharing target data across the mesh network to engage incoming threats like anti-ship missiles after suffering losses.

Yet, as the system grows more autonomous and resilient, the spectre of an uncontrollable AI war machine looms. Scientists and military planners in China are acutely aware of the risks.

Chinese military researchers are conducting rigorous testing to ensure that large-scale weapon systems cannot be triggered automatically – no matter how intelligent the network becomes, according to openly available information.