Xi's WAIC2026 pitch: AI under human control, cities rebuilt around people, and a warning on unequal access

On the morning of 17 July 2026, Chinese President Xi Jinping opened the World Artificial Intelligence Conference in Shanghai with a message calibrated for an audience that has spent two years asking whether Beijing intends to out-build, out-deploy or out-govern the West on artificial intelligence. His answer, delivered in three distinct set-pieces across the day, was to reframe the question. AI, Xi argued, is not principally a race for capability. It is a problem of political choice about who holds the lever.

Between roughly 04:47 UTC and 06:20 UTC, the Chinese leader moved through three arguments that, taken together, sketch a doctrine for a technology Washington and Brussels have so far preferred to discuss in supply-chain and export-control terms. The first: that AI must remain under human control. The second: that the cities AI will reshape must be built around people, not the other way around. The third: that unequal access to the technology risks producing what he called "new historical injustices," a phrase the BRICS-aligned press picked up and amplified within minutes.

The sequencing matters. Capability, urban form, distribution: each is a separate policy lever, and by tying them to a single presidential voice in a single day, Beijing signalled that it intends to compete on the framing of AI governance as well as on the silicon.

The first lever: control as a political claim

Xi's opening call that "AI is always under human control" is the line the Western wire services carried. Read in isolation, it sounds generic: of course human beings should keep their hand on the wheel. Read against the export-control architecture the United States has built around advanced chips since 2022, it reads as a deliberate counter-claim. Washington has framed the frontier of AI as a matter of physical choke points, advanced lithography, leading-edge fabrication, the handful of fabs in Taiwan, South Korea and the Netherlands that can produce the hardware behind frontier models. Beijing's counter is that the choke points are downstream of the choice. If AI is to remain a tool of human political authority, then control is exercised by the state that sets the rules, not the state that fabricates the wafer.

That is a structural argument dressed up as a slogan. It implies that export controls are a contestable interpretation rather than a settled physical reality, and that the legitimacy of AI deployment is a question for sovereign governments rather than for chip designers or model laboratories. The slogan also does work inside China itself, where the Ministry of Industry and Information Technology, the Cyberspace Administration of China and provincial governments have spent two years drafting rules on generative model registration, training-data provenance and algorithmic recommendation. Xi's phrase gives those regulators a presidential anchor without committing to any specific technical standard.

The second lever: cities as the unit of deployment

The second set-piece, on "people-centred urban renewal," was the softer of the three and the easiest to miss. Xi used the WAIC platform to argue that AI should reorganise urban life around citizens, not around surveillance throughput or commercial extraction. The phrasing rhymes with policy work China has pursued for the better part of a decade: the national new-type urbanisation plan, the hukou reforms rolled out in provincial pilots, the smart-city infrastructure that links municipal services to commercial cloud platforms run by Alibaba, Huawei and Tencent.

What the speech adds is the explicit framing that urban renewal is part of the AI agenda, not a separate portfolio. In Western capitals, the smart-city conversation has been dominated by privacy litigation, vendor consolidation and post-pandemic debates over public-space data collection. Beijing's pitch is that the alternative is available and that the choice is political: cities can be built so that residents experience AI as a public utility, with the state as guarantor of equitable access, or as a private extraction machine, with the developer as principal. The pitch is structurally convenient, because China's integrated municipal and industrial policy allows the central government to coordinate housing, transit and AI infrastructure in a way that fragmented Western jurisdictions cannot easily replicate.

The third lever: distribution as legitimacy

The third set-piece is the one BRICS-aligned outlets carried fastest. The phrase "new historical injustices" is doing serious work. By locating unequal AI access on a historical timeline, Xi extends a frame Beijing has used elsewhere, on development finance, on vaccine equity during the pandemic, on the architecture of cross-border payments: that the prevailing international order produces unequal outcomes and that those outcomes compound over time. The argument is that if AI capability concentrates in a handful of jurisdictions and corporate platforms, the result will not be a temporary gap to be closed by diffusion. It will be a structural injustice, written into the next century of economic and political geography.

This is the lever where the political risk is highest. On capability, Beijing can plausibly point to domestic model deployments, to the rollout of Chinese-designed AI assistants in domestic handsets and to the export of surveillance and smart-city packages across the developing world. On urban form, China has a comparative advantage in execution that is hard to dispute. On distribution, the case is harder, because the most restrictive AI export controls in the world today are, in significant part, written by Washington and aimed at Chinese firms. Xi's framing inverts the picture: it casts the United States as the actor entrenching inequality, and Beijing as the actor warning against it.

What the Western wire will and will not carry

A reader of Reuters, Bloomberg or the Financial Times will see the first line and almost nothing else. The "human control" formulation travels; the urban-renewal speech does not, because it is presented as a domestic governance speech rather than a geopolitical marker; the "new historical injustices" line will appear, if at all, as a quote in a round-up paragraph. The shape of the coverage will be: a single slogan, attributed to the Chinese president, with a paragraph on the export-control backdrop and a quote from a Western analyst about the gap between rhetoric and practice.

The Chinese state press will carry the opposite weighting. CGTN and Xinhua will treat the three set-pieces as a single doctrine. Global Times commentary will probably argue, in the days ahead, that Western coverage is itself an exercise in framing, picking the line that flatters Western policy and burying the lines that complicate it. That is a fair objection. Coverage routinely defers to the language of official spokespeople, and when the official in question is the Chinese president, the deference flows in two directions. Western outlets prefer the slogan that fits their existing narrative. Chinese outlets prefer the slogans that fit theirs. The full architecture of what Beijing is proposing requires reading all three together.

What to watch next

Three things are worth tracking in the days and weeks ahead. First, whether the Ministry of Foreign Affairs builds the "new historical injustices" phrase into its BRICS+ and SCO communiqués ahead of the autumn summit cycle, which would signal that the line is policy rather than rhetoric. Second, whether domestic regulators move on the urban-renewal pledge in a way that constrains how commercial AI vendors deploy in Chinese municipalities, particularly the recommendation algorithms that govern content delivery inside city boundaries. Third, whether the United States responds to the framing directly, through a Treasury or Commerce statement, or indirectly, through another tranche of chip controls. Xi's speech is structured to force that response, because the only coherent rebuttal is to argue that export controls are themselves a form of human control exercised legitimately by a sovereign government. That is an argument the Biden and now the post-2024 administrations have so far refused to make on the record.

The uncertainty worth naming is the gap between the speech and the silicon. China has credible capability in application-layer AI, in computer-vision deployment at municipal scale and in industrial robotics. It has less credible capability, on the evidence currently public, at the frontier of foundation-model training, where leading-edge GPUs and high-bandwidth memory remain the binding constraint. The doctrine Xi outlined on 17 July is a coherent politics. Whether the underlying industrial base can carry it into 2027 is a separate question, and one the speech, by design, does not address.

Desk note: Monexus carried all three of Xi's 17 July WAIC2026 set-pieces as a single doctrine, rather than isolating the "human control" line as Western wires did. The urban-renewal and historical-injustice framings are read here as deliberate political architecture, supported by the day's own sequencing rather than as colour.

Wire provenance

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

• https://t.me/bricsnews
• https://en.wikipedia.org/wiki/World_Artificial_Intelligence_Conference
• https://en.wikipedia.org/wiki/AI_safety_in_China