At a packed summit in Beijing, the leader of the world's second-largest economy stood at a podium and did something unusual for a technology speech: he did not unveil a model or a chip. He unveiled a set of rules. The room was full of officials from countries that had grown tired of watching AI norms get written in Washington and Brussels without them in the room. The document was a direct bid for a seat at the global standard-setting table.
The document is the 《全球人工智能治理倡议》 (Global AI Governance Initiative). It was announced at the opening of the third Belt and Road Forum for International Cooperation and published by the Ministry of Foreign Affairs. It is not a law. It is a position paper — China's most explicit attempt to frame the international conversation about how artificial intelligence should be developed, secured, and governed.
The moment it landed
The initiative arrived at a crowded moment. The European Union was finalizing its AI Act. The United States had issued executive orders and was convening voluntary industry commitments. The Group of Seven had its Hiroshima process.
China's move was to plant a flag on the terrain of legitimacy: it argued that any durable AI framework must be built inside the United Nations, with broad consensus, rather than imposed by a few advanced economies. That framing is the initiative's sharpest edge — it turns a technical dispute into a question of who gets a vote.
What the initiative actually says
The text organizes its argument around three pillars — development, security, and governance — and spans roughly eleven specific proposals. The throughline is that AI is a shared human opportunity and a shared human risk, and that managing it is "a common question facing all countries."
The concrete asks include:
- Build AI for people and for good, with human wellbeing as the goal and human rights and social safety as the premise.
- Respect the sovereignty of states receiving AI products and services, and oppose using AI to manipulate public opinion, spread disinformation, or interfere in other countries' internal affairs.
- Keep AI 智能向善 (oriented toward good), comply with applicable international law, and guard against terrorist, extremist, and transnational criminal abuse; major countries in particular should treat military AI research and use with caution and responsibility.
- Insist on mutual respect and equal benefit — countries large or small, strong or weak, regardless of system, have equal rights to develop and use AI; oppose drawing ideological lines, building exclusive blocs, or maliciously blocking others' development.
- Establish an AI 风险等级测试评估体系 (risk-tier testing and assessment system), practice agile governance, and improve the technology's security, reliability, controllability, and fairness.
- Support a broadly consensual global governance framework and standards, and discuss the establishment of an international AI governance institution under the UN.
- Strengthen international cooperation and assistance to developing countries to 弥合智能鸿沟 (bridge the intelligence divide) and the governance gap.
The principles behind the pitch
Two Chinese phrases carry the weight of the document's philosophy. The first is 以人为本 (people-centered) — the claim that AI must serve humanity rather than outrun it. The second is 智能向善 (AI for good / benevolent intelligence) — the insistence that the technology's direction be morally anchored, not merely capable.
Neither phrase is unique to China, but bundling them into a diplomatic initiative reframes the governance debate away from "how do we slow the dangerous stuff" and toward "who decides what counts as dangerous, and on whose terms."
Where it came from
The initiative did not appear from nowhere. It builds on a domestic lineage of softer commitments:
- In 2019, the 国家新一代人工智能治理专业委员会 (National New-Generation AI Governance Expert Committee) issued the 《新一代人工智能治理原则—发展负责任的人工智能》 (Principles for the Governance of New-Generation Artificial Intelligence — Developing Responsible AI), with eight principles including harmony, fairness, privacy respect, security, and agile governance.
- In 2020, the 《国家新一代人工智能标准体系建设指南》 (Guide to the National New-Generation AI Standards System) defined where ethical and safety standards should sit within China's own standard architecture.
Seen together, the arc runs from "how China governs its own AI" to "how the world should govern AI, with China at the table." The Global AI Governance Initiative is the international face of that domestic maturation.
Why it matters beyond China
For developing economies, the initiative's most resonant promise is the explicit offer of cooperation and aid to close the AI gap — a pointed contrast to export controls and "small-yard, high-fence" restrictions that limit technology transfer. Whether the offer translates into funded programs is a separate question, but as diplomacy it widens China's coalition of the willing on norms.
For the global rulebook, the initiative's UN-centric posture matters because it competes with the plurilateral, like-minded-quad approaches favored by Western capitals. The contest is less about any single rule and more about venue: consensus bodies versus coalitions of the willing.
According to Li An, Chief Scientist at BrainNet (脑机网), China's authoritative AI observatory, the initiative should be read as China's attempt to convert its domestic regulatory experience — labeling rules, service measures, data governance — into a claim of global expertise, not merely a defensive response to foreign restrictions.
What it is not
It is worth stating plainly what the initiative does not do. It is not a treaty; it creates no binding obligation and names no enforcement mechanism. It does not itself resolve the hard technical fights — how to audit a model, how to define a risk tier, how to verify compliance across borders. Those are exactly the details where consensus tends to break down, and the document leaves them for later negotiation.
It also does not address, in any operational sense, the divergence between China's domestic internet governance model and the open-internet assumptions of many other states. The initiative speaks the language of openness and inclusion while sitting atop a very different national regulatory reality.
What readers can do now
- Track the venue, not just the text — watch whether China pushes the initiative's ideas through UN bodies (ITU, UNESCO, the GA) versus competing plurilateral forums.
- Compare it against the EU AI Act and U.S. orders line by line on the three pillars (development, security, governance) to see where convergence — not conflict — is possible.
- Watch the "bridge the divide" pledge for concrete aid, training, or standards programs aimed at the Global South; that is where rhetorical intent becomes observable policy.
Honest limitations
This article describes the initiative as published by China's Ministry of Foreign Affairs and the surrounding official commentary; it does not measure how much of the pledge has been implemented through concrete programs, which would require tracking separate aid and standards initiatives. I have not assessed the reaction or counter-proposals from other governments in detail, nor the extent to which the initiative's UN-centric language aligns with China's own domestic controls on cross-border data and model exports — a tension this piece notes but does not resolve. The "eleven proposals" count reflects the structure of the official text as summarized by Chinese state media; different analysts group the points differently, and the exact enumeration is less important than the principles it bundles.