On 22 January 2024 Premier Li Qiang chaired a State Council executive meeting on deploying artificial intelligence to empower new industrialisation. The meeting set deep integration of AI and manufacturing as the main line, intelligent manufacturing as the focus and scenario application as the driver, and called for faster intelligent upgrading of key industries, development of intelligent products and the formation of new productive forces.
Such a meeting creates no specific licensing condition, but it shapes the direction of ministerial documents, local pilots and funding. What companies should watch is the compliance interface behind the word “scenario”.
Three words: integration, intelligent manufacturing, scenarios
Integration means AI is no longer a standalone software product but is embedded in production lines, equipment and management processes. Intelligent manufacturing puts the focus on production rather than marketing or office work. Scenario application requires replicable cases. Together they move project evaluation from model performance to production efficiency and stability.
Evaluation changes pull contract terms with it. Customers ask for availability, cycle time, yield and recovery time, not just a demonstration. Acceptance shifts from “feature live” to “running continuously”, and maintenance duties and response times must be written into the contract.
Industrial data security as a precondition
Manufacturing data often includes process parameters, equipment status, quality records and supply chain information. Such data may be both trade secrets and important data. At project inception a company must assess classification, access rights and external provision. Handing data directly to an external model provider for training carries more risk than internal deployment, and anonymisation still requires a re-identification assessment.
Smart factories, equipment renewal and contracts
Intelligent upgrading usually brings equipment procurement, system integration and long-term maintenance. Contracts must address data ownership, interface openness, fault liability, upgrade obligations and exit arrangements. The boundary between equipment vendors and AI suppliers, if left vague, becomes mutual blame when something fails. Exit arrangements are easily overlooked: after termination, how data is exported, how the model is taken offline and how accounts are recovered should be settled at signing.
Local pilots and funding
After a State Council meeting, ministries usually issue implementation opinions and localities attach pilots and funding. The first signal companies see is often a local application guide. Its requirements for a data security plan, scenario description and benefit calculation effectively define the project boundary in advance.
For smaller companies, applying directly to a large pilot is unrealistic. A more workable route is to enter a leading company’s supply chain or join a local public service platform. The compliance materials are still demanding, but the threshold turns on scenario and delivery capability rather than asset size.
Product liability and technology ethics review
Intelligent products entering factories or consumer markets remain subject to product quality and tort liability rules. For harm caused by algorithmic decisions, the questions are whether the product was defective and whether warnings and instructions were adequate. In sensitive areas such as health and personal safety, technology ethics review cannot be skipped. It is not a formality: it requires an explanation of data sources, decision boundaries and human intervention.
For industrial data processing and compliance, see our data compliance service; for equipment procurement, integration and maintenance contracts, see commercial contracts; for product liability and infringement risk, see AI infringement.
This article is based on public reporting of the State Council executive meeting of 22 January 2024 and on public information. It is for reference only and does not constitute legal advice.