China’s 15th Five-Year Plan Sets National Targets for AI Agent Deployment

Summary

The article says China’s 15th Five-Year Plan for 2026–2030 commits more than ¥1 trillion to AI development, with about 30% reportedly allocated to autonomous agents. It describes sector-specific targets: agent-operated smart factories, nationwide robotaxi networks, broad hospital adoption for diagnostics and surgical assistance, more than 100 million household and office agents, and swarm-agent integration into unmanned defense systems. The plan is presented alongside a Level 4+ autonomy target by 2028 and a requirement that critical-infrastructure agents use China-hosted data and models without foreign dependencies. The article says Chinese companies are already deploying agent systems in logistics, e-commerce, content, recommendations, and factories, while domestic models such as DeepSeek-V3, Qwen-3, and Pangu provide the reasoning layer. It also highlights a reported 600,000-plus agents deployed in 2025, an ambition to reach millions by 2028, exaFLOPS-scale compute for agent training, and a goal of running billion-parameter agent models on edge devices. China is described as competing to establish agent interoperability standards, including Huawei’s A2A-T protocol, while holding 45% of global AI-agent-related patents according to the article’s cited WIPO data. For OpenClaw users, the article argues that model-agnostic architecture and self-hosting could become more valuable as US-China technology ecosystems diverge. It also notes that China’s proposed safety framework would require human involvement for high-risk agents, reinforcing the need for permissions, allowlists, and other guardrails. These claims and targets are presented by the OpenClaw article as the plan’s implications; the supplied text does not independently verify them.