On 25 April 2025 the Political Bureau of the CPC Central Committee held its twentieth collective study session on strengthening the development and regulation of artificial intelligence. Putting “development” and “regulation” side by side in the title of the session is itself an indication of the basic framework of current policy.

The three qualifiers: beneficial, safe, fair

The session called for China’s AI to develop in a healthy and orderly way in a beneficial, safe and fair direction. The three qualifiers map onto value orientation, risk floor and competitive order, and together they form the standard by which the results of AI development are judged.

For industry, “beneficial” means an application must carry value that society can accept; “safe” means risk control is a precondition rather than an after-the-fact remedy; “fair” concerns the availability of factors such as data and compute, and the maintenance of competitive order.

Application orientation and self-reliance

The session also stressed self-reliance and an application-oriented approach. The former points to autonomous control of foundational capabilities such as compute, models and data; the latter points to converting technical capability into value in real scenarios.

Both sit within an existing policy line. See the State Council’s deployment on AI development and the opinion on the AI Plus initiative: on the application side the instruments cluster around opening scenarios, empowering industries and digitalising small and medium-sized enterprises, while on the foundational side they revolve around compute networks and data institutions.

What the new nationwide system means

The session called for giving full play to the advantages of the new system for mobilising resources nationwide. In AI this phrasing usually corresponds to major scientific and technological programmes, compute infrastructure and shared data platforms built under coordinated planning.

The practical significance for companies is this: participants in compute scheduling, public data and industry large models need to understand the logic of coordinated resource allocation rather than relying solely on market competition. See the national integrated compute network service platform.

Where companies should look for the landing points

First, compliance capability is becoming an entry condition. Under a framework that gives equal weight to development and regulation, a company that can clearly explain its data sources, model boundaries and content responsibility is more likely to clear a review. See our notes on generative AI filing in practice.

Second, watch for the follow-on documents. A top-level framing is normally followed by specific documents from ministries and localities, and it is their enforceable provisions that companies need to align with. See our note on the intelligent economy in the 2026 government work report.

Third, turn policy signals into internal decision parameters. The pace of scenario opening, compute coordination and data supply will affect product launch timing and the structure of resource commitments.

The significance of a top-level signal lies not in what new thing it says, but in the stability it confirms for an existing direction.

This article is based on public reporting of the Political Bureau of the CPC Central Committee’s twentieth collective study session on artificial intelligence and public information and is for reference only; it does not constitute legal advice.