Recommended Articles
Legal & Policy

China Drafts Sweeping AI Export Controls — and Opens Them for Comment

By Angeled · November 3, 2026 · Legal & Policy
China Drafts Sweeping AI Export Controls — and Opens Them for Comment
China Drafts Sweeping AI Export Controls — and Opens Them for Comment

China's Ministry of Commerce (MOFCOM) and Ministry of Science and Technology (MOST) issued Announcement No. 48 on October 16, 2026, proposing to add 40 technologies to the country's Catalogue of Technologies Prohibited or Restricted from Export. The draft, analyzed widely on November 2, would significantly expand China's technology export controls — with artificial intelligence accounting for the majority of the new items.

The proposal is a draft, not final law. It is open for public comment until November 15, 2026, and its final form may change. But its scope already tells a clear story about how China intends to govern the outflow of its most advanced technologies.

What would be controlled

Of the 40 technologies proposed, 14 would be prohibited from export and 26 would be restricted, meaning export would require government approval. Twenty-three of the items fall under AI, alongside five in biomedicine and 12 in photovoltaics.

The AI items are strikingly specific. They include model weights and related technology at thresholds of 10^26 floating-point operations of training compute, 2 trillion parameters, or global top-10 standing — thresholds designed to capture only the most capable frontier systems. The proposal also covers AI agents, offensive-cyber AI, compute clusters, and chip design technology, extending controls from models themselves to the tools and infrastructure used to build and deploy them.

These thresholds matter because they define the boundary between ordinary technology trade and strategic capability. By targeting only the largest and most capable systems, the draft aims to control frontier AI without choking off the broader technology trade on which China's economy depends.

What is not covered — and what applies immediately

The draft contains notable exemptions and clarifications. API and software-as-a-service access is reportedly not counted as an export, which would preserve the ability of foreign users to access Chinese AI services remotely even as the underlying technology is controlled. Public-domain technology is exempted in part, as are products made using restricted technology in certain cases.

At the same time, an interim pre-export registration requirement via a MOFCOM platform applies immediately — including to transfers already underway. Companies engaged in covered technology transfers cannot simply wait for the final rules; they must register now. That combination of a draft proposal with immediate interim obligations creates a compliance challenge for any company operating near the thresholds.

Related controls

The October 19 announcements that accompanied the AI proposal also revised controls on drones and tunnel-boring machinery, with product controls effective November 18, 2026. The broader package reflects a coordinated tightening of China's export-control regime across multiple strategic sectors, not an AI-only initiative.

The global context

China's draft arrives amid an intensifying global contest over AI technology controls. The United States and its allies have spent years restricting the flow of advanced chips and manufacturing equipment to China; China's draft represents a reciprocal move, asserting control over the outflow of Chinese AI capabilities. The result is a world in which frontier AI technology faces restrictions in both directions — a significant shift from the relatively open research environment that characterized the field's early years.

For companies operating across borders, the practical consequences are already real. Technology transfer decisions that were once routine now require legal review against multiple, evolving control regimes. The interim registration requirement means that even the comment period is not a grace period.

Why it matters for founders and investors

For founders building AI products with international ambitions, the draft is a planning document as much as a policy proposal. Model scale, parameter counts, and capability rankings are now potential regulatory triggers. Architecture choices — what is exported, what is accessed via API, where compute sits — have compliance consequences that should be considered early, not after a product is built.

For investors, the proposal underscores a structural change in the AI market: the global technology stack is fragmenting along national lines. Companies whose value depends on unrestricted cross-border technology flow face a new category of regulatory risk, while companies that can navigate multiple control regimes — or build within a single one — may gain an advantage. The comment period ending November 15 is the window to understand, and potentially influence, where the final lines are drawn.

Angeled.com connects founders with investors.

Read Next