China’s order blocking Meta Platforms’ multibillion-dollar acquisition of AI startup Manus has turned a Silicon Valley talent-and-technology deal into a broader test of how far Beijing will go to police the overseas transfer of artificial-intelligence assets with Chinese origins.

The National Development and Reform Commission’s foreign-investment security review office said it had prohibited foreign investment in the Manus project and required the parties to withdraw the acquisition transaction. The brief official notice did not name Meta directly, but Reuters reported that the order targeted Meta’s purchase of Manus, a deal valued at more than $2 billion. The agency said the decision was made in accordance with laws and regulations, without providing a detailed public explanation of the legal basis or the specific risks identified.

For Meta, the intervention cuts into one of the company’s most urgent strategic priorities: building stronger AI agents and integrating them across its family of consumer and business products. Meta has spent heavily on AI infrastructure, models and product teams as it competes with OpenAI, Google, Anthropic, Apple, Microsoft and a widening field of startups. Manus was viewed as relevant to that race because its technology centers on agentic AI, a category of software designed to perform multi-step tasks with limited human supervision.

The case is U.S.-relevant not only because Meta is headquartered in Menlo Park, California, but because it raises a practical question for nearly every large American technology company: whether AI acquisitions involving founders, engineers, intellectual property or historical R&D ties to China can close without Chinese national-security clearance. That question matters to corporate development teams, venture investors and founders who have treated Singapore, the United States and other jurisdictions as potential bases for global AI companies with Chinese-born teams or China-developed technology.

Reuters reported that Manus was originally linked to Chinese startup Butterfly Effect, also known as Monica.im, and later moved operations to Singapore. The company had raised funding from U.S. investors and was pursuing international expansion. The transaction, according to Reuters, came under Chinese scrutiny after Meta completed the acquisition, creating uncertainty about how a completed deal could be unwound and what obligations would fall on the companies, investors or executives involved.

The Chinese action is notable because it appears to look beyond the target company’s place of incorporation. In conventional cross-border deal analysis, a buyer often focuses on where the target is legally domiciled, where its shares are held and which merger-control or foreign-investment regimes formally apply. The Meta-Manus dispute suggests Beijing may also examine where core technology was created, where the founding team is based, whether China-based staff contributed to the product, how data and intellectual property were transferred, and whether a restructuring abroad was designed to avoid Chinese oversight.

That broader theory of jurisdiction could significantly change the risk profile for AI startups with China-linked histories. Many founders have used offshore holding companies, Singapore headquarters or U.S. investment vehicles to access global capital and customers. In sensitive areas such as AI agents, model tooling, data infrastructure, chips, cybersecurity and robotics, Beijing’s position suggests those structures may not shield a transaction if regulators view the assets as strategically important.

Meta has said, according to Reuters, that the transaction complied with applicable law and that it anticipated an appropriate resolution to the inquiry. The company has not publicly laid out a detailed plan for how it would respond if China insists that the acquisition be reversed. A forced unwind could be complicated if employees have already been integrated, code has been reviewed or merged, product road maps have changed, or investor payouts have already been distributed.

The block also lands in the middle of a widening U.S.-China struggle over AI supply chains. Washington has restricted China’s access to advanced semiconductors and has increased scrutiny of outbound investment in sensitive Chinese technology sectors. Beijing, in turn, has tightened control over technology exports, data transfers and transactions it sees as involving national security. In that context, the Manus case is not simply a single acquisition dispute; it is part of a larger contest over whether frontier AI talent and intellectual property can move freely across borders.

For Silicon Valley, the immediate implication is that speed may no longer be enough in AI M&A. Large technology companies have often moved quickly to acquire small AI teams, secure researchers, buy product traction or prevent rivals from gaining access to scarce talent. But if an acquisition involves a company that was founded in China, built core technology there or relocated only recently, buyers may need to treat Chinese regulatory review as a closing condition rather than a remote political risk.

The deal could also affect startup financing. Venture firms backing AI companies with Chinese-origin teams may now face tougher diligence from later-stage investors and strategic buyers. Questions that once sat in legal footnotes—where was the IP created, who owns the models, where do engineers sit, what data was used, were any technology-export approvals required—may become central to valuation and exit planning. A startup that cannot clearly document its ownership of technology and compliance history may find that a future sale to a U.S. platform carries a discount.

The case is especially important for agentic AI because that market is moving quickly from demos to enterprise workflows. AI agents promise to automate research, coding, customer support, sales operations, software testing and internal business processes. For Meta, agentic systems could support advertising tools, business messaging, creator products and consumer assistants. For regulators, the same capabilities may raise concerns because they combine software autonomy, data access and potentially transferable know-how.

China’s subsequent tightening of rules around overseas deals, technology transfers and sensitive sectors has added to the message. Reuters reported on June 1 that Beijing issued new rules aimed at overseas transactions involving Chinese investors, technology, data and national security, with provisions that could formalize state power to review or reverse certain transactions. That broader framework makes the Meta-Manus dispute look less like an isolated intervention and more like an early signal of a hardened policy stance.

There are still important uncertainties. Public Chinese notices have been brief, and Meta has not provided a detailed public account of the transaction structure, the operational status of Manus inside Meta, or the specific steps it may take in response. It is also unclear how Chinese regulators would enforce an unwind against entities or assets now outside mainland China, or whether a negotiated remedy could preserve part of the transaction while satisfying Beijing’s concerns.

Still, the direction of travel is clear enough for U.S. technology companies: AI dealmaking is becoming a national-security exercise, not just a product or talent acquisition strategy. For Meta, the Manus block threatens a specific bet on AI agents. For the broader Silicon Valley ecosystem, it creates a precedent that may slow, complicate or reshape the next wave of cross-border AI acquisitions.