China Officially Releases Three Industrial AI National Standards, Co-Drafted by SUPCON

SUPCON (中控技术), the Chinese process-industry automation vendor, helped draft three national standards for industrial artificial intelligence that have now been officially released and put into effect, according to a report carried on Sohu News on August 31.

The three documents are GB/Z 204-2026 "AI Industrial Large Model System Architecture," GB/Z 203-2026 "AI Industrial Large Model Technical Requirements," and GB/Z 195-2026 "AI Industrial Agent Reference Architecture." Together they form China's first systematic standardization framework for industrial AI, filling a gap in the field's guiding standard system.

The pain points behind the standards go beyond technology alone. Industrial large models previously lacked a unified architecture and technical specification, with fragmented construction across platforms and scenarios, while industrial agents had no unified guidance suited to complex plant sites, making it hard to replicate innovations across production lines and enterprises.

GB/Z 204-2026 proposes a complete "six-in-one" architecture covering infrastructure, data, key technologies, models, applications, and security protection, running through the full lifecycle of industrial large-model development, adaptation, deployment, and application for standardized management from planning to operations.

GB/Z 203-2026 sets technical requirements across pretraining, adaptation and fine-tuning, inference, and deployment over the full lifecycle, mutually reinforcing the architecture standard to guide standardized development and commercial deployment of industrial large models.

The most notable item is the industrial agent reference architecture. Developed domestically with no equivalent standard internationally, it is the most forward-looking of the three, defining the reference architecture of industrial agents from a functional perspective and clarifying core components including large and small model libraries and eight functional modules — perception, memory, reasoning, planning, decision-making, execution, communication, and evolution — plus two-way interaction with industrial physical space and information systems.

SUPCON became one of the enterprises deeply involved in writing the standards, drawing on its industrial AI expertise and proven results across nearly 200 installations at process-industry leaders including PetroChina, Sinopec, Wanhua Chemical, China Coal, and Huayi Group.

With top-level architecture, technical rules, and agent frameworks now in place, a complete "framework-method-deployment" standard system is taking shape. SUPCON says it will keep advancing industrial AI innovation on the plant floor and promote the standards across more industries, working with the wider ecosystem to push industrial AI toward standardized, systematic, and scaled development.