China rallies world for AI rules
China's AI action plan positions as AI rule-shaper at a time when the U.S. deregulates and the EU tightens controls.
Publishes 13-point AI governance plan on 26 July 2025, defining artificial intelligence as a “global public good” and pushing for international cooperation.
Seeks open-source collaboration and UN-aligned safety standards within 12 months, with Shanghai proposed as a global AI governance hub.
Commits to sustainable AI, including energy benchmarks, low-power chips, and global standards to reduce environmental impact.
Signals confidence despite funding dip: while Asia's AI funding fell one-third in H1 2025, China's generative-AI user base now exceeds 250 million.
On July 26, 2025, China unveiled its most comprehensive AI policy to date—the Artificial Intelligence Global Governance Action Plan at the World Artificial Intelligence Conference in Shanghai.
The launch followed Premier Li Qiang’s speech urging collaborative AI development and proposing the formation of a global AI cooperation organization. This plan follows the 2023 Compute Infrastructure Plan and aligns with China’s call to position AI as both a growth engine and a global public good.
Blueprint compresses 13 pledges into one roadmap
Beijing's policy states that:
Artificial intelligence is a new frontier of human development and must be developed to serve humanity.
Compute capacity surges toward 300 EFLOPS
China's AI drive is grounded in infrastructure.
The 2023 Compute Infrastructure Plan set a target to exceed 300 EFLOPS of compute power by 2025, with 35% allocated to smart computing.
The government continues to issue "compute vouchers" and fund AI mega projects to accelerate this expansion.
With continued US export controls on Nvidia's most advanced GPUs, demand is rising for local chips such as Huawei’s Ascend and Biren’s BR100, both central to national AI scaling.
Open-source pivot targets global alignment
Clause five of the plan calls for "safe and reliable open-source platforms" and the development of international communities for shared innovation.
The aim: reduce duplication, improve interoperability, and accelerate real-world applications.
The strategy is already bearing fruit. Open-source model DeepSeek surpassed 100 million users after releasing its base model to developers.
Chinese LLM developers now open-source at a pace unmatched by U.S. counterparts, both to avoid overseas licensing friction and to secure wider international adoption.
Energy, data, and safety protocols address global concerns
China’s new action plan recognizes AI's growing carbon footprint. Clause seven introduces a "sustainable AI" mandate: develop energy- and water-efficient chips, promote green computing practices, and create enforceable performance standards.