TL;DR

China has barred Manus AI co-founders Xiao Hong and Ji Yichao from leaving the country while regulators review Meta’s $2 billion acquisition of the AI agent startup. The founders were summoned to Beijing, questioned about potential foreign investment violations, and told they cannot travel internationally. Meta shares dropped 3% on the news. This is the first time China has used exit bans to block a major AI acquisition — and it won’t be the last.

What Happened

On March 25, reports from the Financial Times and Reuters confirmed that Chinese authorities have imposed exit bans on two Manus co-founders. Xiao Hong, the CEO, and Ji Yichao, the chief scientist, were summoned earlier this month to a meeting in Beijing with the National Development and Reform Commission (NDRC) — China’s most powerful economic planning body.

The pair were questioned about whether Manus violated foreign direct investment rules when it relocated from China to Singapore in mid-2025. After the meeting, both were told they could move freely within China but could not leave.

No formal charges have been filed. No official investigation has been opened. But the message is clear enough.

The Deal That Started It All

Meta announced the acquisition of Manus in December 2025 for roughly $2 billion, making it Meta’s third-largest acquisition ever, behind WhatsApp and Scale AI.

Manus had rocketed to prominence after launching what it called “the world’s first fully autonomous AI agent” in March 2025. The product could handle multi-step tasks (market research, coding, data analysis, content creation) inside a sandboxed virtual computer environment. Think of it as giving an AI its own desktop to work at, complete with a browser, terminal, and file system.

Within eight months of launch, Manus hit $100 million in annual recurring revenue. The company employed about 105 people across Singapore, Tokyo, and San Francisco.

For Meta, the acquisition was about catching up in the AI agent race. While OpenAI, Google, and Anthropic all had agent products in the market, Meta’s own agent efforts were lagging. Buying Manus gave them a ready-made platform and the team that built it.

Why China Is Blocking It

Here’s the core issue: Manus was built in China. The core IP, the early team, the initial research all happened in Beijing under the entity “Beijing Butterfly Effect Technology.”

Throughout 2025, Manus systematically relocated. The headquarters moved to Singapore. State-linked investors were replaced with US venture capital, including a $75 million Series B led by Benchmark. By the time Meta came knocking, Manus looked like a Singapore company on paper.

Chinese regulators see this differently. They’re examining whether Manus effectively exported sensitive AI technology without the required government approvals. The relocation from Beijing to Singapore, followed by the sale to a US company, looks to regulators like a deliberate end-run around China’s technology transfer rules.

Analyst Dai Menghao from law firm King & Wood Mallesons put it bluntly: the AI agent technology Manus developed is “definitely something that Chinese regulators could subject to export controls.”

The “Singapore Bath” Problem

The Manus situation exposes a strategy that’s become common among Chinese AI founders: build the tech in China, incorporate in Singapore, sell to Western buyers. Industry watchers call it the “Singapore bath”: a quick offshore restructuring designed to wash away the regulatory complications of Chinese origin.

It worked for a while. Several Chinese AI startups have used this playbook to attract Silicon Valley funding and avoid the geopolitical minefield of US-China tech relations.

China just signaled that the bath water has run cold.

By barring the Manus founders from leaving, Beijing is establishing a precedent. The message to every Chinese AI founder considering a similar move: your technology started here, and we have leverage over you personally, regardless of where your company is incorporated.

Wendy Chang from the Mercator Institute for China Studies framed it as a deliberate strategy: “Beijing’s probe appears designed to prevent Chinese AI technology and talent from being absorbed by foreign acquisitions, particularly in the United States.”

What This Means for Meta

The immediate impact is operational paralysis. Xiao Hong and Ji Yichao are physically separated from Meta’s headquarters and the engineering teams they’re supposed to integrate with. You can’t run an acquisition integration over video calls when the two key leaders are stuck in a country where they’re under regulatory scrutiny.

Meta’s official response has been measured. A spokesperson told Reuters: “The transaction complied fully with applicable law. We anticipate an appropriate resolution to the inquiry.”

Behind the scenes, Manus is actively seeking law firms and consultancies in China to help resolve the situation. But there’s no clear timeline, no established process to follow, and no guarantee of any particular outcome.

Meta shares fell about 3% on the news. The $2 billion price tag is a rounding error on Meta’s $1.5 trillion market cap, but the signal matters more than the dollar amount. If Meta can’t close AI acquisitions with Chinese-origin teams, that narrows the talent pool significantly.

The Bigger Picture: AI as a Strategic Asset

The Manus situation goes beyond one deal. It marks a shift in how governments treat AI capabilities: commercial products are becoming strategic assets, right up there with semiconductors and defense technology.

The US has been tightening the screws on AI chip exports to China for years. Companies like Tesla are even building their own chip fabs to reduce dependence on Asian supply chains. The January 2026 BIS rule moved from blanket denials to case-by-case reviews for chips like the NVIDIA H200, but next-gen Blackwell chips remain blocked. Congress advanced the AI OVERWATCH Act with a 42-2 vote. Export control enforcement budgets jumped 23%.

China is now playing the same game in reverse. If the US controls the hardware supply chain, China is asserting control over the talent and model supply chain. Exit bans on founders are the human equivalent of chip export restrictions.

Murthy Grandhi from GlobalData summed it up: “The battleground has moved from semiconductor chips to AI models, agents, talent, and enterprise deployment. Controlling the flow of talent is as critical as controlling hardware or software infrastructure.”

What Comes Next

The Manus situation will likely play out in one of three ways:

Scenario 1: Conditional approval. China allows the deal to proceed with conditions, perhaps requiring that certain technology stays in China, or that Manus maintains a Chinese subsidiary with continued access to the IP. This is the outcome Meta is probably hoping for.

Scenario 2: The deal collapses. If China determines that the technology transfer violated export controls, the acquisition could be unwound entirely. Meta would lose the technology and the team. Manus would be stuck in regulatory limbo with founders who can’t travel.

Scenario 3: Prolonged freeze. The most likely outcome in the short term. No formal charges, no resolution, just indefinite uncertainty while both sides try to find a face-saving exit. The founders remain in China. Integration stalls. The technology ages.

For the rest of the AI industry, the implications are already being felt. Expect fewer clean exits for Chinese-origin AI startups selling to US buyers. Expect more due diligence on the geographic origin of AI IP. Expect the “Singapore bath” strategy to get a lot harder.

Winston Ma from NYU School of Law called this a potential test case for China’s equivalent of CFIUS (the Committee on Foreign Investment in the United States). If the precedent holds, every future cross-border AI deal involving Chinese-origin technology will need to account for this risk.

FAQ

Can the Manus founders still work on the product?

Technically yes. They can communicate with Meta remotely and work from within China. But effective acquisition integration requires physical presence, and both founders are separated from Meta’s global engineering teams. The practical impact on product development is significant.

Has China done this before with tech companies?

Exit bans on business executives are not new in China, but applying them specifically to block an AI acquisition is unprecedented. China has used similar tactics in financial fraud investigations and tax disputes, but the Manus case represents a new application of this power to technology transfer disputes.

Does this affect other AI startups based in Singapore?

Yes. Any AI company with Chinese-origin technology, founders, or IP that has relocated to Singapore or another jurisdiction should be reassessing their exposure. The “Singapore bath” strategy that many startups relied on has been directly challenged by this action.

Limited, at best. China’s NDRC has broad authority over foreign investment reviews, and there is no established appeals process for exit bans in this context. Meta’s best path is likely diplomatic and commercial negotiation rather than legal challenge.

Could this kill the deal entirely?

It’s possible but not certain. Meta has signaled it still wants the acquisition to proceed. The more likely near-term outcome is an extended freeze while both sides negotiate. If China formally determines that export controls were violated, unwinding the deal becomes a real possibility.

Bottom Line

China just drew a line. If your AI technology was built on Chinese soil, by Chinese engineers, with Chinese research — Beijing considers it a strategic asset, no matter where your company is incorporated today.

For Meta, this is a $2 billion lesson in geopolitical risk. For the broader AI industry, it’s a preview of the next decade: every major AI deal will now be a three-party negotiation between buyer, seller, and the governments that claim jurisdiction over the technology. The era of clean, fast AI acquisitions across the US-China divide is over.