China's NDRC: AI Industrial Policy and the Blocked Manus Deal

A planning agency, not a regulator in the Western sense

The National Development and Reform Commission sits under China's State Council and functions as the country's top macroeconomic planning body, tracing its lineage to the 1952 State Planning Commission. It drafts and coordinates China's five-year plans, approves major investment projects, and sets prices and industrial policy across sectors, a mandate that spans energy, infrastructure, price controls and, increasingly, the digital economy. [2]

That breadth is why the NDRC turns up in AI industrial policy despite having no dedicated technology regulator's brief. It also administers foreign investment security review, the authority it later used against Meta. [2][6]

"AI Plus" and the commission's role in diffusion

In 2025 the State Council issued an "Artificial Intelligence Plus" action plan directing AI to be integrated across the economy and society by 2035. The NDRC, as the government's macro planning agency, published an accompanying explainer describing the goal as elevating AI from a specific application into infrastructure that other industries build on, and it is the agency coordinating how that integration is tracked and funded across sectors rather than a body that sets safety standards for AI systems themselves. [3]

Atlas interpretation: That framing matters for reading the Manus decision below: the NDRC's interest in AI is chiefly economic and strategic, about where capability and talent sit and who controls them, rather than about content moderation or model safety, which other Chinese bodies handle. [3]

Blocking the Manus acquisition

On April 27, 2026, the NDRC ordered Meta and Manus to unwind Meta's roughly $2 billion acquisition of the Singapore-based, Chinese-founded AI agent startup. Meta had structured the deal as a purchase of the Manus "project," its technology and team, rather than the Chinese corporate entity, and the NDRC's order addressed that same distinction: it treated the underlying technology and the engineers who built it as subject to Chinese jurisdiction regardless of where the deal was booked. [4][5]

Legal analysis of the order describes the NDRC invoking China's 2021 Measures for the Security Review of Foreign Investment, in what practitioners characterized as the first publicly confirmed use of that mechanism to reverse an AI acquisition already carried out. By June 2026, Meta had begun separating shared data and operations to comply. China's State Council followed in July 2026 with a new Regulation on Outbound Investment giving the government an ongoing supervisory mechanism over similar cross-border technology deals. [6]

Atlas interpretation: The order reads as Beijing applying to AI talent and technology the kind of outbound control Washington has applied to chip exports: not stopping money from moving, but stopping capability, in this case an agent-building team and its work, from leaving Chinese jurisdiction through a corporate structure. It is the NDRC's economic-planning and security-review authority converging on a single acquisition, rather than a new AI-specific rule. [6]