Industry News

SHANGHAI, July 20 (China Economic Net) ¡ª Global AI governance had a busy week. Days after the United Nations held its first Global Dialogue on AI Governance in Geneva, 29 countries signed the founding agreement of the World AI Cooperation Organization (WAICO) in Shanghai, where the World Artificial Intelligence Conference (WAIC) also produced a run of governance outcomes: a UN industrial AI governance framework developed with Chinese and British institutions, initiatives on interoperability and trust among AI agents, an international action plan on AI ethics, and Chinese pledges of training and cooperation for developing economies, among many others.

Behind it all, China has spent years writing and enforcing domestic AI rules, at a scale no other country has attempted. How much of that experience can be shared ¡ª and what would other countries actually take?

What is there to share ?

The answer starts with what global governance now needs. For most of the past decade, the work consisted of agreeing on principles ¡ª that AI should be fair, transparent, safe, subject to human oversight. "Some of those principles have pretty much settled down. There's a convergence around common principles pretty much everywhere in the world," John Higgins, chair of the International AI Governance Association, told China Economic Net (CEN) on the sidelines of WAIC. "Now we're seeing the work to build on those principles."

That means instruments rather than statements: labels a user can see, certificates an auditor can issue, test reports a regulator can accept. Distilling principles into practical, implementable deliverables is "not easy, but it can be done," Higgins said, and is already under way in international standards bodies ¡ª the venues, he noted, where engineers from companies across different countries work on the same technical problems.

At an AI governance forum held on the sidelines of WAIC on Saturday, speakers set out four kinds of instrument that China's large-scale deployment has produced.

The first is certification. Huawei received ISO/IEC 42001 certification this March ¡ª the first international standard for AI management systems, setting requirements for how an organization governs AI across its life cycle ¡ª covering five core business groups and nearly 100 sites worldwide, audited and issued by the Switzerland-based body SGS. Gan Bin, Huawei's vice president for standardisation and industry development, told the forum the company wants governance requirements standardized globally, so that "when we build products, there are rules to follow."

The second is content marking. Under rules in force since September 1, 2025, AI-generated text, images, audio and video must carry a visible notice plus machine-readable metadata embedded in the file, so platforms can still identify synthetic content after it has been copied or re-uploaded; a national standard specifies how those marks are written. Zhang Chao, an associate professor of electronic engineering at Tsinghua University, described the technical chain now being built around that requirement.

The third is sector testing. In the auto industry, where AI shapes both supply chains and driving decisions, Fan Hailong of China Merchants Testing Vehicle Technology Research Institute described compliance built into the development cycle itself ¡ª factory testing, over-the-air updates, maintenance records ¡ª with audit trails kept in a form that can establish responsibility after an incident. An intelligent vehicle, he said, is "a moving AI terminal."

The fourth concerns the environmental cost of AI itself. Zhang Jian, deputy dean of Tsinghua's Institute of Climate Change and Sustainable Development, said sustainable development has been written into WAICO's core objectives as a "sustainable AI" concept: prioritising green power for AI workloads, reducing and recycling water used in data-center cooling, and managing hardware through its full life cycle.

How might it travel?

At the AI governance forum, Zhu Xufeng, dean of Tsinghua's School of Public Policy and Management, described two routes. The first is policy diffusion: governments drafting their own rules study what China has done and borrow selectively, absorbing it gradually rather than adopting it wholesale. The second is multilateral co-creation on problems no country can handle alone, where frameworks are written jointly.

Both routes are selective by design ¡ª governments take what fits ¡ª which matches how the demand side describes its own needs.

For developing countries, their priorities in AI governance may begin with digital infrastructure, then regulatory capacity, then skills, Higgins told CEN. On the middle item, he said, countries without the means to build a framework from scratch "can take ready-made solutions from others that suit them."

China's announcements in Shanghai tracked that sequence: 5,000 AI training places for developing countries over five years, international AI application cooperation centres to be built with ASEAN, the Arab League, the African Union, CELAC, the SCO and BRICS, and the deployment of "Mazu," China's AI-powered meteorological early-warning system, in 30 countries.

But developing countries can also contribute. Speaking to CEN on the sidelines of WAIC, Jason Slater, Chief AI, Digital and Innovation Officer of the UN Industrial Development Organization (UNIDO), said the Global South holds "a huge amount of untapped talent and data," and that the unsolved problems there are themselves an asset ¡ª a source of the use cases and regulatory sandboxes that turn principles into tested practice. "We should not underestimate the desire of such countries to make use of AI and to help craft a future dialogue around governance," he said.

A start, with work to do

China "already contributes hugely in global AI governance," Higgins told CEN, through its research institutions and its role as a convener.

The hurdles ahead are practical ¨C certificates recognised abroad, watermarks readable across platforms, test reports trusted by foreign regulators. The UN's 2024 report Governing AI for Humanity identified three persistent gaps in global governance: representation, coordination and implementation.

An early case may come from Mongolia, which has been explicit about wanting to build with Chinese partners. Former president Nambaryn Enkhbayar told the forum his country plans to send young Mongolians to take graduate degrees in AI so it can develop the technology in cooperation with countries such as China, and that it wants to work with Tsinghua University and Chinese companies on a digital reconstruction of Karakorum, the ancient Mongol capital. Slater responded on the spot, inviting Mongolia to join the first pilots of UNIDO's new framework. This might be what the first mile of Chinese AI governance looks like on its journey abroad.