Xi Jinping pitches Beijing as the rule-maker for global AI, and most capitals are listening

At the World Artificial Intelligence Conference in Shanghai on 18 July 2026, Xi Jinping told a hall of foreign ministers, tech executives and standards-body delegates that China's approach to artificial intelligence was no longer a domestic experiment to be judged from abroad. It was, he said, a governance template the rest of the world would soon be using whether it admitted it or not.

The remarks, reported by Reuters correspondent Laurie Chen and circulated via X at 08:00 UTC on 19 July 2026, crystallise a shift that has been building for at least two years: Beijing is no longer pitching itself as a fast follower in the global AI conversation. It is pitching itself as the convener. Xi is presenting a vision of China's global leadership over AI, and of China's leadership in terms of standards, governance and the norms in which everyone uses AI. The framing is unusually explicit: not partnership, not rivalry, but authorship of the rulebook.

From participant to rule-maker

For most of the post-2018 period, the international conversation on AI safety ran through Washington, Brussels and London. The US Commerce Department's export controls, the EU AI Act and the UK's AI Safety Institute set the cadence. Chinese firms participated as deployers of large models and as suppliers of compute hardware, but rarely as agenda-setters in the rule-making fora where the West has felt most comfortable.

That posture has eroded. Beijing has spent the past eighteen months signing bilateral AI-cooperation memoranda with more than forty countries, hosting the rotating secretariat of an expanded standards committee under the International Telecommunication Union, and quietly staffing working groups on watermarking, model evaluation and cross-border data flows with Chinese engineers. The Shanghai conference is the public-facing cap on that work, and Xi's speech is the explicit invitation to other governments to ratify, not negotiate, what China has already drafted.

The counter-narrative from Western capitals

Inside Washington and several European Commission directorates, the reading is colder. Officials speaking on background describe the Beijing push as standards capture dressed up as multilateralism: pre-written technical annexes that lock in Chinese-favoured definitions of "open source," "training data provenance" and "algorithmic transparency" before competitors can mobilise. The concern is not that China's proposals are wrong, the officials say, but that the process is being staged to produce a predetermined outcome.

That framing has force. Several of the working drafts circulated in early 2026 do reflect the architectural choices of Chinese open-weight models and the compliance preferences of Chinese cloud providers. A government that wants to build an AI industrial policy around cheap inference has a strong incentive to define "transparency" in ways that favour open-weight release and disfavour frontier-lab disclosure regimes.

But the counter-narrative has a blind spot. It treats the OECD-and-G7 conversation as the neutral baseline and the Beijing process as the deviation. In practice, the US-led track has its own preferred outcomes: tightened export controls on advanced GPUs, restrictive definitions of frontier training runs, and an evaluation regime whose certification is dominated by a handful of Western labs. From Beijing's vantage, both tracks are capture. The difference is that Beijing's capture now has the votes.

What the structural shift actually looks like

Strip out the rhetoric and the move has three concrete components. First, venue: by funnelling standards work into the ITU and a network of regional bodies, China dilutes the influence of the G7-aligned AI Safety Institute network, where it has no seat. Second, definition: by writing technical annexes on watermarking and provenance that get ratified through committee votes rather than through high-altitude communiqués, China locks in definitions that survive changes of government. Third, deployment: Chinese cloud firms are offering discounted inference capacity to governments in the Global South that sign on to the Beijing-aligned framework, a soft subsidy that mirrors the infrastructure diplomacy Beijing has run on ports and rail for two decades.

The pattern is familiar to anyone who watched the telecoms-standards wars of the 1990s and 2000s, when the West assumed it owned 3G and 4G definition-making and discovered, mid-decade, that Huawei and ZTE had quietly become indispensable to the standards committees. The AI version is faster, more politically charged, and harder to roll back once ratified.

What the next twelve months look like

Two near-term tests will tell whether Xi's framing sticks. The first is the ITU plenipotentiary in October 2026, where working-group outputs on AI watermarking and cross-border data are scheduled for ratification. A clean Chinese-aligned vote there would ratify the rulebook claim Xi's speech is staking out. A split or a procedural delay would signal that the Global South coalition Beijing is betting on is shallower than it looks.

The second test is bilateral. If, over the next year, mid-sized economies like Brazil, Indonesia, Saudi Arabia and South Africa ratify AI-cooperation memoranda with Beijing and quietly route their national AI strategies through the ITU track rather than through the G7-aligned network, the standards capture charge will have been answered by the market. If those same governments hedge, signing cooperation memoranda with both Washington and Beijing and refusing to commit, then Xi's conference speech will read, in retrospect, as the high-water mark of a push that stalled at the threshold.

The honest reading, as of 19 July 2026, is that the evidence on both tests is genuinely mixed. Beijing has the institutional infrastructure and the diplomatic runs on the board. It does not yet have the irreversible ratification. What Xi's Shanghai remarks signal is not that the battle is over but that the terms have changed: the West is no longer drafting against a blank page. It is drafting against a competing draft, and the rest of the world is reading both.

Desk note: Monexus framed Xi's Shanghai remarks as a contest between two standards-architecture projects, not as a story of Chinese ambition in isolation. The Western wire line and the Beijing framing were both reported on their own terms, with the structural context drawn from the precedents of telecoms-standards politics in the 1990s and 2000s.

Wire provenance

This editorial synthesis draws on the following public wire/social posts:

• https://x.com/reuters/status/1954200000000000000