Agentic AI is not the same as embodied AI, but the article linked in today's Daily Planet gives us some insight into how China is thinking about the former.
China is rapidly advancing in the field of "agentic AI," which refers to artificial intelligence systems capable of performing complex tasks and acting on behalf of users. Several Chinese companies, including DeepSeek, Tencent, Alibaba, and ByteDance, have developed impressive AI models and frameworks that go beyond simple chatbots to take autonomous actions. For example, Manus, developed with Tencent's support, and AutoGLM-Rumination from Zhipu AI, have demonstrated state-of-the-art performance in this area.
In Chinese discussions, the term "agentic AI" is expressed using various words that reflect slightly different meanings. One common term is 代理 (dàilǐ), meaning "agent" or "proxy," emphasizing AI acting on someone's behalf. Another set of terms, like 自主代理 (zìzhǔ dàilǐ) or 智能体 (zhìnéngtǐ), highlight the AI's autonomy and independence as an intelligent actor. This variety shows that the exact definition of agentic AI in China is still evolving as the technology matures.
The Chinese government is actively shaping the development and regulation of these AI systems. The China Academy of Information and Communications Technology (CAICT) has released standards and reports focusing on the technical capabilities, safety, reliability, and controllability of agentic AI. These efforts align with broader government goals to lead in AI innovation while ensuring ethical use and risk management. For instance, CAICT's standards guide how AI agents should be developed and evaluated, and international cooperation is also encouraged.
Overall, China's approach to agentic AI combines rapid technological progress with careful policy planning. While formal laws specifically targeting agentic AI are still emerging, existing regulations on data, cybersecurity, and AI applications already apply. The government emphasizes balancing innovation with safety, aiming to build trustworthy AI systems that can operate autonomously but remain under human control.