In January 2026 the Ministry of Industry and Information Technology, the Cyberspace Administration of China and six other ministries jointly issued the Implementation Opinion on the “AI Plus Manufacturing” Special Action, to accelerate the application of AI in manufacturing. It is the sectoral implementation of the broader “AI Plus” initiative.
Quantified targets to 2027
The Opinion sets out that by 2027 China’s key AI technologies will be supplied reliably and securely, and includes a set of quantified targets: three to five general large models deeply applied in manufacturing, 100 high-quality industrial datasets, 500 typical application scenarios, and a number of ecosystem-leading enterprises with global influence.
The logic is clear: the focus shifts from technology breakthroughs to industry adoption. Whether a large model can enter manufacturing depends less on parameter scale than on matching industry-scenario data and replicable application patterns.
Data is the real bottleneck
The target of 100 high-quality industrial datasets points to the core constraint: data. Industrial data is scattered across equipment, production lines and information systems, and involves trade secrets and production safety; its collection, aggregation and circulation are more complex, compliance-wise, than consumer scenarios.
This aligns with the National Data Administration’s work on industry high-quality datasets — see our note on the high-quality datasets action plan and the National Data Administration’s dataset guide. For manufacturers, the barrier to joining dataset construction is not technology but clarity of data ownership and authorisation chains.
Compliance touchpoints for manufacturers
First, classification and grading of industrial data. Whether production data, process parameters and equipment-status data constitute important data must be judged by sectoral requirements. Where personal information is involved (such as employee behaviour data or facial-recognition attendance), it must be assessed separately. See our note on network data security risk assessment measures.
Second, filing duties for models and algorithms. Services with public-opinion attributes or deep-synthesis capability require filing; internal use is mainly governed by data and security rules. See our algorithm filing service.
Third, contractual arrangements with upstream and downstream parties. Joint R&D, data sharing and model deployment involve multiple parties; IP ownership and data-use boundaries must be clear in contracts.
Fourth, dual-track assessment for manufacturers going global. Industrial AI products for the EU market may fall within high-risk systems. See our note on the EU AI Act transparency rules taking effect.
The other side of the policy dividend
Special actions bring subsidies, pilots and scenario access — and greater compliance visibility. Joining government-organised pilots usually carries data-security and algorithm-compliance review. Companies should prepare compliance materials at the application stage rather than catching up at acceptance.
Competition in manufacturing AI is shifting from “whose model is stronger” to “who can embed the model safely into the production line.” Policy has set the direction; the gap will show up in the details of data governance and compliance design.
This article is based on the Implementation Opinion on the ‘AI Plus Manufacturing’ Special Action issued by eight ministries and public information. It is for general reference only and does not constitute legal advice.