The four-agency stack that shapes how AI products launch in China, the mechanisms each one enforces, and the operating tempo that fuses compliance with product development.
Western coverage treats Chinese AI governance as one faceless "state." The operational reality is a four-agency stack with distinct mandates and a release tempo measured in days.
The deeper point: in China, product, compliance, and distribution ship as one bundle, not three. This card is the org chart that replaces the Cold War analogy.
Each agency owns a distinct lever — content, telecoms, public security, market — and each one ships an engineering deliverable on the same cadence as the product itself.
In China, product, compliance, and distribution ship as one bundle. The four-agency stack does not slow the release calendar — it sets it.
It is the build. Algorithm filing, security assessment, and content moderation ship as engineering deliverables inside the same sprint as the model itself. Western analysts often miss this because their mental model of regulation is post-hoc: build the product, then submit for approval, then launch.
China's leading AI labs do not work that way. Filing requirements, security review checklists, and content moderation policies sit on the same Jira board as the training pipeline. The compliance team is not downstream of engineering. It is engineering.
The result is a release tempo measured in days, not quarters. What looks from outside like state interference is, from inside the firm, a stable and predictable engineering constraint that has been priced into the product roadmap from day one.
When China's generative AI rules took effect in August 2023, Baidu reached public launch in sixteen days.
The speed did not reflect regulatory absence. It reflected a system where filing, security review, and product development had merged into a single sprint years earlier. The book is built around scenes like this one.
If you cover China AI, brief boards on it, or build models that touch its market — three operating instructions from this card.
"Wild West copycat" and "Soviet top-down monolith" are equally wrong. Model the four-agency stack instead, with distinct deliverables per agency.
They sit on the same Gantt chart. An exposure model that treats them as separate variables tends to underprice the moat by a structural margin.
The CAC registry is the leading indicator of which models reach 100M+ users. Press cycles trail it by 30 to 90 days.
This card explains the governance machinery. The book traces how that machinery emerged, how it shaped the country's response to U.S. export controls, and how it now exports along with the technology itself.
China's AI dominance began as commercial necessity, not a top-down research project. Baidu and others built terabyte-scale ML systems to serve search and online-to-offline super-apps.
The Wei Zexi scandal forced the shift from unregulated growth to the four-agency Regulator Stack. Algorithm filing became a mandatory engineering deliverable. This is the act this card is drawn from.
U.S. chip export controls forced radical efficiency: DeepSeek's Mixture-of-Experts, FP8 quantization, and accelerated adoption of Huawei's Ascend domestic hardware.
A permissive consumer environment, aggressive price wars, and massive business-to-government procurement pipelines drove generative AI adoption from 2023 to 2025 at unprecedented scale.
Chinese firms bypass domestic constraints by exporting open-weight models and bundled infrastructure to Southeast Asia, the Middle East, and Africa — exporting governance norms alongside the technology.
A mechanism-level operational manual for the world's second-largest AI ecosystem. August 2026.
Built from primary Chinese-language regulatory texts, company filings, and technical documentation. Drawing on the CAC algorithm filing registry, MIIT licensing publications, MPS cybersecurity grading standards, SAMR enforcement decisions, and corporate disclosures from Baidu, Alibaba, ByteDance, Tencent, and DeepSeek.
