China’s National Development and Reform Commission ordered Meta to reverse its more than $2 billion acquisition of Manus, the agentic AI startup. The NDRC’s foreign investment security review office issued a single sentence: it had decided to “prohibit foreign investment in the Manus project in accordance with laws and regulations” and required both parties to withdraw the transaction. No elaboration was offered. None was needed.
The decision ends a four-month regulatory process that began almost immediately after Meta announced the deal in December 2025. Manus, originally developed by the Chinese company Butterfly Effect, had relocated its headquarters and most operations to Singapore before the sale. Meta had already begun integrating roughly 100 Manus employees into its Singapore offices. CEO Xiao Hong had a direct reporting line to Meta’s chief operating officer. The integration was, by most accounts, going well. That turns out not to have mattered.
The exit bans are the detail worth sitting with. In March, co-founders Xiao Hong and Ji Yichao were summoned to Beijing for questioning by the NDRC and subsequently barred from leaving mainland China. Both had been living in Singapore. Chinese regulators were examining whether moving Manus personnel and technology from China to Singapore, then selling to a U.S. company, constituted an unauthorized transfer under China’s export control framework, which was expanded in 2020 to cover certain algorithms. The regulatory term of art here is “Singapore-washing.” It is not a compliment.
The restrictions on Manus’s founders fit a pattern that has been assembling itself quietly over the past year. In early 2025, Chinese authorities began advising the country’s leading AI researchers and entrepreneurs to avoid travel to the United States. DeepSeek founder Liang Wenfeng did not attend an AI summit in Paris in early 2025, a period during which Chinese authorities were advising leading AI figures to limit international travel. Staff at DeepSeek’s parent company, High-Flyer, have reportedly surrendered their passports, with the company restricting international travel for employees working on core AI models. The combination of travel advisories and passport collection does not have a single name yet, but the pattern is legible enough.
These measures sit alongside a softer apparatus. The phrase “investing in human capital” has appeared repeatedly in China’s 2025 government work report and long-term planning documents. More than one-fifth of academic programs have been adjusted in the past two years to meet demand in fields the government considers strategically important. Local governments have introduced housing subsidies and equity incentive programs aimed at attracting and retaining technology talent. Carrots and sticks, deployed simultaneously.
One detail is easy to miss in the geopolitics. Manus is not a foundation model company. It is an agent orchestration platform, a runtime layer that turns large language models into autonomous systems with access to tools, browsers, and code execution environments. Beijing’s intervention suggests that this capability layer, not just model weights or training data, now falls within the scope of what regulators consider worth protecting. The controlled-technology aperture just got wider.
Lian Jye Su, chief analyst at the technology research group Omdia, said the decision shows China “is willing to play hardball when it comes to AI talents and capabilities, which the country views as a core national security asset.” For founders who were considering the Singapore relocation playbook, the message is now quite clear.
Sources: TechCrunch, CNBC, CNN
–
By the Control Plane Editorial Team