China's PLA AI Procurement: Drone Swarms and Deepfake Tools on the Table
A PLA procurement officer’s hand pointing at a line item for 'deepfake generation software' in a printed document, surrounded by stacks of identical procurement forms under studio lighting, highlighting the bureaucrat...📷 AI illustration
- ★Thousands of PLA procurement documents analyzed
- ★AI for drone swarms and deepfake tools
- ★Autonomous decision-making systems in development
A trove of procurement documents from China's People's Liberation Army lays out a clear intent: weaponize AI at scale. Researchers at Georgetown University's Center for Security and Emerging Technology analyzed thousands of these requests, finding concrete mentions of drone swarms, deepfake generation tools, and autonomous decision-making systems. The full analysis on The Decoder shows that Beijing is not just exploring—it's actively funding and developing these capabilities.
But there's a critical difference between a procurement request and a field-ready system. The documents describe experiments and prototypes, not mass deployment. Drone swarms in a controlled test range behave very differently from swarms operating under electronic warfare, weather variability, and contested communications. The same applies to deepfake tools—generating a convincing video in a lab is one thing; deploying it in real-time psychological operations at scale is another entirely.
From procurement requests to battlefield deployment: the gap is still wide
A single autonomous decision-making system prototype mounted on a test drone chassis inside a sterile clean-room lab, surrounded by tangled sensor cables and diagnostic monitors, highlighting the gap between procureme...📷 AI illustration
The Deployment Gap
The real challenge for China's military AI push is bridging the gap between demonstration and operational reality. Autonomous decision-making systems, for example, require robust sensor fusion, fail-safe logic, and trust in unpredictable combat environments. None of that shows up in a procurement request. As noted in the Georgetown study, the PLA's modernization roadmap includes three overlapping phases: mechanization, digital networking, and intelligent warfare. We are likely still in the transition between phase two and three.
Hardware and Safety Limits
Hardware constraints also remain. Swarm drones need reliable low-latency links; deepfake tools need massive compute and training data; autonomous systems need redundant control loops. These are not solved by procurement alone. The research brief emphasizes that while the scale of investment is large, specific budgets and timelines remain undisclosed. Until we see these systems in sustained field exercises, the procurement documents are a signal of intent, not capability.