Manus Exit Ban: When AI Founders Become Export-Controlled Assets
China bars Manus AI co-founders from leaving the country as regulators probe Meta's $2B+ acquisition. From TikTok's forced sale to Manus's exit ban, US-China tech controls escalate from corporate entities to individual founders.
Announced Dec. 30, 2025
to signed deal
Xiao Hong (CEO) + Ji Yichao (CSO)
during Singapore relocation
Key Findings
What happened: China has imposed exit bans on the co-founders of AI startup Manus, marking an escalation from controlling corporate entities to restricting the physical freedom of individual founders. This is a turning point in export control history — people themselves have become the controlled asset.
The symmetry: The US used PAFACA legislation to force TikTok to divest Chinese ownership. China is using export control reviews to block Manus from being sold to an American company. Mirror images, asymmetric tools — the US deployed legislation, China deployed administrative restraint.
Core judgment: When a state begins restricting personal freedom to control the flow of technology, the signal extends far beyond any single deal. For China's AI startup ecosystem, this is not an enforcement action — it is a rewriting of the rules.
I. The Deal: Ten Days to Close, Three Months in Limbo
On December 30, 2025, Meta announced its acquisition of Manus, a Singapore-based AI agent company, for over $2 billion. The deal took just ten days from first contact to signed agreement — a timeline that speaks plainly to Meta's urgency. Manus had reached a $125 million annualized revenue run rate in under three years, and its general-purpose AI agent product, launched in March 2025, briefly seized global attention, earning comparisons to DeepSeek. CEO Xiao Hong was to become a Meta vice president reporting directly to COO Javier Olivan, with the roughly 100-person team joining Meta intact.
Less than two weeks after the announcement, Beijing responded. On January 8, 2026, MOFCOM spokesperson He Yadong announced a review of the acquisition, covering export controls, technology import and export administration, and outbound investment compliance. In March, the National Development and Reform Commission (NDRC) summoned Manus co-founders Xiao Hong and Chief Scientist Ji Yichao for questioning, after which both were barred from leaving mainland China — free to move domestically, but unable to travel internationally.
A deal struck in ten days is now trapped in a review that will last at least ninety. The real question, however, is not whether this transaction survives, but what the emerging regulatory logic behind it reveals.
- 2022Xiao Hong founds Butterfly Effect Technology in Beijing, launches AI browser plugin Monica
- Mar. 6, 2025Manus launches general-purpose AI agent product, drawing global attention
- Apr. 2025$75M Series B led by Benchmark at $500M valuation; Tencent, ZhenFund, HongShan Capital participate
- Mid-2025Relocates HQ from Beijing to Singapore, shuts down China operations, lays off 80 mainland staff, goes dark on Chinese social media
- Dec. 30, 2025Meta announces acquisition for over $2B, negotiated in approximately 10 days
- Jan. 8, 2026MOFCOM announces review of the acquisition, covering export controls, tech transfer, outbound investment
- Mar. 2026NDRC summons Xiao Hong and Ji Yichao; both barred from leaving China; probe expands to cross-border capital flows and tax accounting
II. The Mirror: TikTok and Manus, Same Logic, Opposite Directions
Place the Manus affair alongside TikTok's forced divestiture and a symmetrical structure comes into sharp focus: the US forced a Chinese company to sell; China is preventing a Chinese company from being bought. Two governments severing the same cord that once connected their tech markets.
In April 2024, Congress passed the Protecting Americans from Foreign Adversary Controlled Applications Act (PAFACA), requiring ByteDance to divest TikTok's US operations or face a nationwide ban. In January 2025, the Supreme Court upheld the law's constitutionality. By December, TikTok reached a joint venture agreement with Oracle, Silver Lake, and other American investors, reducing ByteDance's stake below 20 percent and nominally transferring algorithmic control to the US side. The joint venture was formally established in January 2026.
Manus traces the exact reverse trajectory: a company born in China, whose core technology was developed in China, first moved its registration to Singapore, then sold itself wholesale to an American tech giant. Beijing's response was to summon the founders, launch a multi-agency investigation, and bar key personnel from leaving the country.
| Dimension | TikTok (US Intervention) | Manus (China Intervention) |
|---|---|---|
| Direction | Force Chinese company to sell US operations | Block Chinese company from being acquired by US firm |
| Legal instrument | Congressional legislation (PAFACA) + Supreme Court endorsement | Administrative action (NDRC summons + MOFCOM review + exit ban) |
| Control target | Corporate entity (ByteDance's US subsidiary) | Natural persons (founders Xiao Hong, Ji Yichao) |
| Stated rationale | National security (data sovereignty, foreign adversary control) | Export control (technology transfer, FDI compliance) |
| Actual objective | Sever Chinese influence over US user data and information flows | Prevent Chinese AI core technology and talent from flowing to the US |
| Outcome | ByteDance forced to cede control, stake reduced below 20% | Deal in limbo, founders cannot leave, integration stalled |
The legal frameworks differ entirely, but the underlying logic is identical: amid accelerating US-China tech decoupling, neither side will allow the other to acquire its core technology assets through commercial transactions. The US used its legislative and judicial apparatus to execute what amounted to a lawful expropriation — ByteDance lost control of TikTok's US operations, but the transaction at least proceeded within a legal framework. China's approach was more direct, and rougher: no congressional debate, no court ruling — one meeting with the NDRC, one exit ban, and the deal was paralyzed.
The divergence in methods is not merely a difference in governance style; it is a difference in where the costs fall. The American approach exacts its price in procedural length — nearly two years from PAFACA's passage to the joint venture's establishment. China's approach exacts its price on individuals — two founders' personal freedom became a bargaining chip on the negotiating table.
III. The Escalation: When People Become Export-Controlled Assets
The most significant aspect of the Manus case is not the fate of the deal itself, but the escalation in the tools of control. Traditional export control logic targets goods — chips, source code, algorithm models — using licensing regimes to restrict the cross-border flow of technology. In the Manus case, the object of control shifted from goods to people. Xiao Hong and Ji Yichao were barred from leaving the country, which in essence treats AI leaders as "technology carriers" subject to export control.
This shift did not come without precedent. Since 2023, China has informally restricted semiconductor talent from taking positions in the US. The 2024 revision of the Regulations on Technology Import and Export Administration expanded the list of restricted technologies. But prior controls always targeted corporate conduct — the question was whether a company had improperly transferred technology, not whether a founder could board a plane. The Manus case broke through that boundary: neither founder has been charged with a crime or formally placed under investigation. They were restricted solely because of an ongoing "assessment."
This is restraint without charge. Its deterrent power lies precisely in its ambiguity — no one knows when the "assessment" will conclude, no one knows where the compliance boundaries lie, and no one knows who the next founder to be summoned might be.
The Singapore relocation strategy itself is telling. In mid-2025, Manus shuttered all Chinese operations, laid off 80 mainland employees, deleted its Chinese social media accounts, and completed its sale to Meta as a "Singapore company." Meta's acquisition announcement specifically emphasized that Manus had "no continuing Chinese ownership interests." Beijing's response amounted to a declaration: the registration can move, but the people cannot. So long as the core technology was developed in China and the founders are Chinese citizens, jurisdiction does not evaporate with a change of address.
IV. The Chill: After the Rules Are Rewritten
For China's AI startup ecosystem, the Manus case sends a signal that runs far deeper than a single frozen deal. It effectively invalidated a rule that had been taken for granted: that Chinese AI companies could achieve "compliant internationalization" through structural adjustments — relocating registration, replacing investors, transferring IP. Manus proved that this pathway fails. What was framed as going global was, in Beijing's eyes, closer to flight.
The chilling effect will propagate along two channels. First, any Chinese AI company involved in a cross-border acquisition will now need to incorporate a "political risk premium" into its valuation model — investors will no longer calculate solely on the basis of technology and market size, but also on the probability that founders may be barred from leaving the country. This is not a hypothetical risk; it is a demonstrated reality. Second, AI talent with international ambitions will confront a choice that did not previously exist: building a company in China means your physical freedom may at any point become a variable in state-level negotiations. Once this perception takes hold, its impact on talent mobility is irreversible.
Beijing's own regulatory logic contains a paradox. The purpose of the exit ban is to prevent AI technology from flowing abroad, but the fear generated by such restrictions is itself accelerating the decisions of the most capable AI talent to leave while they still can. The tighter the controls, the stronger the incentive to leave; the stronger the incentive, the tighter the controls need to become. This cycle is accelerating the disconnection between China's AI industry and global innovation networks.
V. The Verdict
TikTok and Manus — two cases, two directions, one conclusion: the window for US-China tech transactions is closing. The US used legislation to lock Chinese companies out of controlling American platforms. China used administrative measures to lock AI talent and technology inside its borders. The two locks together form a de facto bilateral blockade.
What makes the Manus case exceptional is what it reveals about the direction of China's control apparatus — from regulating companies to regulating people, from approval-based systems to restraint-based systems. This is a larger question than whether any single deal survives. When a government employs its citizens' freedom of movement as an instrument of technology control, the message extends beyond any one company. It is a statement about the position of the entire entrepreneurial class: your freedom is conditional, and the condition is the national interest.
Xiao Hong and Ji Yichao saw the risk coming in mid-2025, which is why they moved the company to Singapore as quickly as they could. What they underestimated was a simple truth: in China, a company can leave. People cannot. The price of that lesson will be borne by China's entire AI industry.
读到这里,说明你关注真正重要的事
锐报每周深度分析直送邮箱——财经、地缘、科技,穿透表象。零广告,零废话。