China upgrades its surveillance network with newer AI tools

In short: China’s local police are upgrading camera networks with newer AI that can search video using text and send alerts about certain behaviors.

What's going on

A Financial Times investigation found that local governments across China are buying and deploying upgraded surveillance systems that use more advanced artificial intelligence. The FT said it reviewed more than a dozen public procurement documents and also spoke with people familiar with the contracts.

The goal is to modernise a network that was largely built in the mid-2010s and to move toward “predictive policing”, meaning police try to spot risks earlier instead of responding after something happens. An expert quoted by the FT said the older system was more reactive and less able to interpret intentions.

The new systems use features like computer vision (software that “looks” at video and tries to understand what is happening) and large language models, or LLMs (the type of AI behind many chatbots). One example in the report is text-based video search, where an operator can type a request like “a woman wearing a red hat” and the system finds matching footage. This is like using search in your photo library, but across city camera feeds.

Procurement examples cited include a tender in Yaodu town in Sichuan province that budgeted Rmb900,000 for 175 high-definition cameras with “intelligent video analysis”, and a project in Yancheng, Jiangsu, described as a Rmb7mn platform that would analyse multiple data sources, including social media and camera footage.

What to watch

Human rights groups quoted by the FT warned that better AI could increase the state’s ability to monitor people at scale. Another key point is cost, since newer AI-enabled cameras can be much more expensive, so some places may upgrade servers and software instead of replacing every camera.

Source: Financial Times