New State Council Rules Make AI Deployment Pass an Employment Test

Many expected a document to slow AI down. What arrived instead is a construction blueprint for using AI well. On October 23, China’s State Council General Office published the Opinions on Adapting to AI Development and Fostering an Employment-Friendly Mode of Development (Guoban Fa [2026] No. 29, dated October 16) — eight sections and eighteen measures that put “employment-friendly” directly in the title. This is not a document that restrains AI; it installs a social interface on every AI deployment.

1. Not a Stop Sign, but a Speed Governor

The general requirements are carefully balanced: actively strengthen the positive effects of AI in creating jobs, empowering traditional roles, raising job quality and stimulating entrepreneurship, while effectively mitigating its displacement impact. Both sides of the ledger sit in one sentence — the technology keeps running, but its pace and application scenarios are to be managed. The “AI Plus” action plan continues: eastern regions host core AI industries while central and western regions take on data annotation and smart-device manufacturing, widening the employment funnel first.

2. Three Provisions That Will Change Deployments

Among the eighteen measures, three will alter corporate behavior most.

First, the order of adoption is now written down. Article 6 directs AI to be applied first in understaffed and high-hazard roles, while in labor-intensive industries its scope, intensity and pace must be handled prudently. AI should take the jobs nobody wants before moving toward the jobs everybody has.

Second, projects must pass an “employment gate”. Article 7 requires that AI application projects in market-access domains with potentially major employment impact undergo employment risk assessments alongside project review. Evaluation forms now carry a new column: how many roles displaced, how many transformed, and who signs off.

Third, buffers and notice become encouraged practice. The document encourages companies to explore an unemployment buffer period, to notify affected workers in advance of large-scale AI rollout, and to provide transition support. Buffer length, notice timing and retraining handover will become standard sections of any deployment plan.

3. The Other Hand Builds New Jobs

Reading this file as purely defensive would be a mistake. Section Two is entirely about creation: an AI SME entrepreneurship support program built on open-source ecosystems, state investment funds backing AI ventures through market-based mechanisms, and four entrepreneurship models linking research achievements, industry growth, vocational skills and livelihood needs to startups. Occupational catalogs are expanding too — digital managers, AI engineers and AI trainers are named for accelerated new-occupation standards, with local governments encouraged to add corresponding title-review tracks. Policy is pulling on both ends: moderating the pace of displacement while widening the pipeline of new roles.

4. New Questions for Employers and Workers

For enterprises, rushing no longer guarantees savings. A project with significant displacement effects must answer three questions: who signs the assessment, how long the buffer lasts, and who receives workers after retraining. If the answers are not in the proposal, the project may slow down on the review table. For workers, the risk of a single-skill portfolio is now stated in policy language — the “Skills Light the Way” training campaign, project-based and order-style training, and AI education across all schooling stages are ammunition for transition; injury-insurance pilots for flexible and new-form employment keep expanding.

5. NineZenith Take

Putting “employment-friendly” in the title of a State Council document signals that policymakers now treat AI’s social interface as infrastructure. For NineZenith Technology, which delivers AI to government and enterprise clients, this is not merely a compliance cost — it is a product requirement. Deployment plans should ship with employment impact assessments, human-machine collaboration design and retraining packages, and the value proposition of agents should be rewritten: not replacing people, but taking over hazardous and understaffed work. Technological progress and employment stability were never a single-choice question; this document prescribes the construction order — build new roles first, move old ones second, keep a buffer, and give early notice.

(Based on Guoban Fa [2026] No. 29 published on gov.cn, October 23, 2026.)