In Western debates China is often portrayed as a single, top‑down AI power where Beijing both directs technological development and sets the rules. A study published on September 19 in the Journal of Contemporary China argues that this depiction misses key features: Chinese AI governance is the product of ongoing bargaining and competition among the central state, local governments and major technology firms.
About the study
The paper is authored by Xinhua Fu and Tao Huang. Xinhua Fu is an associate professor at Beijing Jiaotong University Law School and head of its Data Law Research Center, with research on data flows and AI regulation. Tao Huang is an assistant professor at the City University of Hong Kong Law School, working on constitutional law, cyberlaw and platform governance. The authors rely on Chinese‑language laws, local regulations, corporate documents and reputable media reports to demonstrate a more layered regulatory reality than the monolithic‑state narrative suggests.
The centre: setting security and political red lines
The study does not deny strong central control. Beijing primarily defines national security, political and ethical red lines, and sets minimum safety expectations. At around the same time as the paper’s publication, Chinese authorities released an AI Safety Governance Framework 3.0 document that focuses on risk categorization, technical responses and comprehensive governance measures — illustrating the central government’s emphasis on risk management.
Local governments as the accelerator: competition and subsidies
Functionally, the authors argue, China’s AI governance resembles a de facto federalism: although China is not a federal state politically, provincial and municipal governments actively compete to build AI ecosystems. While the centre draws the safety floor, local governments serve as engines for innovation, industrial development and investment promotion.
The paper contrasts Shenzhen’s and Shanghai’s 2022 local rules. Shenzhen’s 2022 regulations adopt an application‑ and market‑oriented approach, encouraging firms to enter the market quickly and providing regulatory sandboxes and access to data and computing resources. Shanghai’s 2022 measures are more systemic, treating data, compute and algorithms as distinct elements and placing greater emphasis on risk management. The authors also highlight local financing: according to the study, Shanghai mobilized roughly 1 billion yuan in support via coupons for compute, models and data sources.
Tech firms are not merely implementers
The study stresses that major Chinese technology firms — such as Huawei, Baidu, Alibaba, SenseTime and Xiaomi — cannot be reduced to passive executors of state will. These companies participate in standard setting, industry initiatives, testing zones, self‑regulation documents and consultations with authorities.
Three illustrative areas are given: autonomous driving, generative AI and smart‑car safety. In autonomous driving, Huawei and Baidu help shape industry safety norms and data‑handling practices as well as develop technology. For generative AI, firms like Alibaba and SenseTime have developed internal safety, ethics and risk management principles for their large language models, which can later serve as references in certification and compliance processes. Following the March 29, 2025 fatal crash of a Xiaomi SU7 vehicle, the ministry in charge of industry urged full testing of combined driving‑assistance systems, clearer functional boundaries and avoidance of misleading marketing; firms adjusted messaging and product positioning toward more precise “assisted driving” claims.
These examples indicate companies influence regulatory outcomes through technical standards, testing and adaptation rather than through overt political contestation.
Soft law, sandboxes and industrial policy
A third key point is that China does not rely solely on a command‑and‑control regulatory model. The governance toolkit also includes soft law instruments, industry standards, regulatory sandboxes, local subsidies and corporate self‑regulation. Thus Beijing acts not only as a regulator but also as an infrastructure builder, risk manager and industrial policy coordinator.
Relevance for the Trump–Xi talks
The September 24, 2026 state visit by Donald Trump and Xi Jinping will address broad issues such as chips, export controls, rare earths, data security, AI safety, industrial competition and strategic trust. The study’s authors argue that a more accurate understanding of China’s internal AI governance matters for diplomacy: seeing China only as a monolithic, top‑down AI state risks underestimating local governments’ incentives and corporate regulatory agency. Conversely, treating China merely as a pragmatic market actor risks overlooking central political and national‑security red lines.
The study does not suggest strategic tensions between the US and China have disappeared. Rather, it recommends that US policymakers distinguish where conflict is structural and where limited, technical cooperation might be possible. China’s AI system, the paper concludes, is a layered regulatory machine composed of central control, local industrial competition and corporate adaptation — a configuration that makes China no less a strategic competitor but explains better how the system operates internally.
Key dates and figures cited in this article:
- Study publication: Journal of Contemporary China, September 19, 2026.
- Trump–Xi state visit in Washington: September 24, 2026.
- Xiaomi SU7 fatal crash: March 29, 2025.
- Shenzhen and Shanghai local AI regulations: 2022.
- Shanghai local support package: approximately 1 billion yuan (per the study).
- Contextual US documents: America’s AI Action Plan and the White House National AI Legislative Framework presented in March 2026.
