China’s new LLM-powered ‘AI brain’ automates satellite surveillance - Interesting Engineering

Chinese aerospace researchers have introduced a new system that leverages Large Language Models (LLMs), or AI, to advance satellite-based surveillance.

The Air Target Agent System integrates large language models (LLMs) with collaborative AI agents. It enables autonomous analysis and decision-making from the sky.

The new development was unveiled amid ongoing debates about the ethics of using AI—and minimizing human decisionmaking—in warfare.

China’s new AI surveillance system

The new system might sound like something a Bond villain would use to spy on entire populations from the sky. The unveiling also comes as the US is facing scrutiny for its use of AI targeting systems in the ongoing Iran conflict.

American systems reportedly automate stages from imagery review to strike selection, though specific details remain classified. Incidents, such as a February strike on an Iranian primary school that killed over 200 children, have intensified debates about AI’s potential for commiting warcrimes, as well as the issue of accountability in AI-assisted warfare.

However, according to a report from the South China Morning Post (SCMP), China’s system represents an explicit effort to boost transparency over its use of AI.

The new system uses AI agents to break down complex tasks, select appropriate algorithms, coordinate workflows, and recover from failures independently. Unlike traditional image recognition tools, it goes further by interpreting satellite imagery, drawing conclusions, and initiating responses without human input.

Researchers describe the architecture as a “brain plus tool army,” where the LLM serves as a central coordinator directing specialized AI tools.

This setup reportedly reduces analysis time significantly while maintaining reliability in dynamic scenarios. In testing, the system demonstrated the ability to handle obstacles autonomously, a key advantage for real-time intelligence and targeting applications.

China’s AI defense push

The technology builds on China’s broader push in AI for defense, where satellite data is increasingly linked to machine reasoning.

“In the future, we will further explore deployment and optimisation strategies in larger-scale, real application scenarios,” Wang Lei, who is affiliated with the Chinese Academy of Sciences, explained, as per the SCMP report.

However, autonomous systems raise similar ethical and operational questions as those levied against the US. China has so far emphasized human oversight in its military AI, but the Air Target Agent System emphasizes minimal intervention.

Though deployment details are scarce, the focus on scalability suggests preparation for integration into larger networks of satellites and command systems.

The development of the new system highlights the intensifying competition in military AI. As both major powers advance automated targeting, the risks of escalation and miscalculation grow.

Analysts expect further refinements as China tests the system in expanded operational settings. The platform also illustrates how AI agents are shifting from supportive tools to core components of modern surveillance and targeting architectures.