China - AI Open Innovation Platforms

Overview

The Guidance on the Construction of National New Generation Artificial Intelligence Open Innovation Platforms, issued by the Ministry of Science and Technology (MOST) and dated August 1, 2019, establishes a national programmatic framework for identifying, building and supervising open AI innovation platforms that can accelerate research‑to‑application pathways across China's economy. The Guidance emphasizes an "enterprise‑led, open and shared" model intended to concentrate resources on clearly defined AI subfields and to provide shared services such as model libraries, algorithm packages, dataset services, and testbeds for developers and enterprises. The text is available through the Chinese central government publication of the notice (MOST notice on gov.cn) and has been translated and summarized by independent policy organizations (CSET translation).

Definitions

The Guidance defines key terms operationally for program implementation: "open innovation platform" refers to an organized, public‑facing R&D and service infrastructure led by a leading AI enterprise (often in cooperation with research institutes and universities) that focuses on a specific AI subfield and provides shared technical resources (models, datasets, APIs, middleware, testbeds). "Leading enterprise" denotes a company demonstrating core technical competency, stable R&D capability and ability to recruit partners. "Open and shared" denotes services made accessible to external developers and SMEs via standardized interfaces, while maintaining compliance with national data and cybersecurity rules. The Guidance also distinguishes "pilot zones" (geographic innovation areas) from platform entities and positions platforms as key inputs into pilot zone development.

Governance and Institutional Framework

The Guidance places MOST at the center of national governance for these platforms. MOST issues the Guidance, prescribes application and evaluation procedures, approves platforms for national recognition, coordinates multi‑ministerial support, and conducts periodic performance reviews. Provincial and municipal science and technology departments are responsible for facilitating local implementation, supporting candidate platforms, and reporting to MOST. Platform governance is expected to be enterprise‑led with participation by academic and research partners; enterprises must establish internal governance mechanisms for operations, intellectual property, and data management. For official publication and administrative details see the MOST/gov.cn notice (MOST notice on gov.cn) and MOST explanatory materials and diagrams (MOST infographic and explanation), which set out roles and responsibilities for each institutional actor.

Key Focus Areas

The Guidance enumerates priority tasks and capabilities that approved platforms must pursue: (1) Focus on well‑defined AI subfields tied to major application needs (e.g., computer vision, speech, autonomous systems, industrial AI). (2) Establish publicly accessible model and algorithm libraries and modularized software components to accelerate downstream development. (3) Provide standardized APIs, interoperability protocols, and middleware to reduce integration costs for SMEs and research teams. (4) Build open testbeds and benchmarking systems for performance evaluation and for safe experimental deployments. (5) Support data sharing (subject to data security and personal information rules) by providing cleaned, anonymized datasets or secure computation environments for testing. (6) Participate in national standards development and contribute to common technical frameworks. (7) Cultivate developer ecosystems through documentation, tutorials, competitions and incubation services. (8) Facilitate industrialization pathways by connecting demonstration projects with city and provincial pilot initiatives. The Guidance also expects platforms to support measurement and validation capabilities (benchmarks, evaluation tools) that feed into national R&D agendas.

Implementation Framework

Implementation proceeds through an application, selection and supervision cycle coordinated by MOST and local science and technology authorities. Candidate platforms should be proposed by leading enterprises in conjunction with institutional partners and local governments; applications document technical capability, collaborative arrangements, infrastructure, and plans for open services and commercialization. MOST evaluates candidates against specified criteria (leadership, technical strength, infrastructure, collaboration network, openness commitments) and designates national platforms. Designated platforms gain policy support opportunities (project funding, inclusion in pilot zone programs) but are also subject to ongoing reporting and performance assessment. MOST retains authority to disqualify platforms or withdraw recognition where performance or compliance is unsatisfactory.

Monitoring and Evaluation

The Guidance requires periodic monitoring and multi‑dimensional evaluation: technical performance (R&D outputs, model libraries, benchmarks), service delivery (APIs, user uptake by SMEs and developers), industrial impact (deployment cases, economic contribution), governance (open governance mechanisms, IP arrangements), and compliance (data protection, cybersecurity measures). MOST and local departments carry out scheduled reviews and may require corrective actions, enhanced reporting, or changes in governance. Evaluation results inform continued funding, recognition and positioning within national pilot zone strategies.

Penalties, Liability, and Appeals

While the Guidance is administrative in nature, it articulates enforcement measures within MOST’s supervisory remit: platforms that fail to meet conditions may be removed from the national roster, lose eligibility for MOST funding or pilot support, or be publicly named. MOST also reserves the right to order remediation or suspend support if non‑compliance with technical or management requirements is identified. Where violations implicate data protection, cybersecurity, or other laws, platforms may be subject to sanctions under separate statutory regimes (e.g., Data Security Law, Personal Information Protection Law, cybersecurity regulations) administered by relevant authorities. The Guidance signals administrative remedies rather than criminal provisions and functions alongside statutory regulatory frameworks that define liability and appeal rights under Chinese administrative procedure rules.

Relationship to Other Instruments

The Guidance implements and operationalizes the State Council’s New Generation Artificial Intelligence Development Plan (国发〔2017〕35号) as a sectoral instrument focused on platform construction. It complements MOST’s subsequent pilot zone guidance (国家新一代人工智能创新发展试验区建设工作指引, August 29, 2019) and feeds into national projects under the Science and Technology Innovation 2030 agenda. The Guidance also interacts with national standardization roadmaps, AI ethics guidelines, and legal instruments on data security and personal information protection. For cross‑reference, see the State Council AI development plan and MOST pilot zone guidance on the central government site (State Council New Generation AI Development Plan (2017) and MOST pilot zone guidance (2019)).

International Alignment

The Guidance is primarily domestic and programmatic but reflects global practice in promoting open platforms, interoperability and shared testbeds. While it does not formally adopt external legal standards, it encourages platforms to engage with international standards and cooperation where appropriate, and to contribute to shared technical norms. Platforms participating in cross‑border activities must also align with China’s data export and cybersecurity rules. Analysts note the Guidance’s parallels with other jurisdictions’ policies favoring public testbeds and open repositories but emphasize that Chinese implementation proceeds within national regulatory constraints on data and security. Relevant English‑language translations and policy analyses are available from independent research centers (CSET translation).

Implementation Timeline

Sources and References