Offshore Shell, Onshore Roots: Jurisdiction in the Meta–Manus Transaction

Introduction

On April 27, 2026, the Office of the Working Mechanism for the Security Review of Foreign Investment, which sits under China's National Development and Reform Commission, issued a decision. It prohibited the foreign acquisition of the Manus project and ordered the parties to unwind the transaction.[1]

On its face, this was a transaction that took place outside China. The buyer was Meta Platforms, Inc. and its affiliates, a company incorporated in Delaware and listed in the United States. On the sell side, the entity that appeared in public was Butterfly Effect Pte. Ltd. of Singapore (Unique Entity Number 202330764R). That entity operates and develops Manus, a general-purpose AI agent. On December 29, 2025, Manus announced that it was "joining Meta." It said that it would continue to operate from Singapore.[2] The press estimated the deal at roughly US$2.5 billion, but the parties never disclosed a formal price.[3]

Both ends of the transaction, one in the United States and one in Singapore, appeared to sit outside China. Yet it was China that ultimately halted the deal and ordered a return to the prior state. This raises a question that cannot be avoided. A company is incorporated and headquartered abroad. On what basis can China regulate it, and on what basis can China make that regulation stick?

This paper uses Manus to answer two questions, on two levels.

The first is a concrete question, and it is the focus of this paper. How is a company's "nationality" determined? When a company is registered offshore but rooted onshore, how strong must the "genuine connection" be before a state is entitled to intervene? The Manus decision provides an important, though incomplete, illustration of this emerging regulatory approach.

The second is a background question, addressed briefly at the end. As more states extend their regulatory reach across borders to intervene in cross-border transactions, these rules overlap. What kind of conflict does that overlap produce?

As to method, this paper prioritizes official and primary materials and uses reputable media reports only for facts unavailable in official records. It distinguishes confirmed facts, reported facts, and analytical inferences.

Part One: The Target and the "Nationality" Problem — How China Characterized the Deal and on What Basis It Asserted Jurisdiction

I. What Manus Is: Sorting the "Project" Out in Chronological Order

To understand what China did, one must first see clearly what was being acquired. Tracing the Manus "project" in chronological order makes the thread clear.

The earliest chapter is the founder's entrepreneurial prehistory. Around 2015, Xiao Hong founded Wuhan Nightingale Technology and built tools such as Yiban and Weiban. That chapter is not the same legal entity as the Manus that came later, but it explains the provenance of the founding team. The starting point was inside China.[4]

In April 2022, Beijing Butterfly Effect Technology Co., Ltd. was established. According to the Wall Street Journal, the early version of Manus was developed by engineers at this company. As of early 2025, founder Xiao Hong still held 28% of it. This was the onshore starting point of Manus's research and development.[5]

Around 2023, Butterfly Effect Pte. Ltd. of Singapore was set up, before Manus's formal release. Google's app store lists it as the developer of Manus, with Unique Entity Number 202330764R, and the company states that it operates from Singapore.[6] Media reports also mention that a Cayman holding layer may exist offshore, used to receive financing and to hold upper-tier equity. Its exact name cannot be verified in the available public materials, and this paper makes no finding on it.[7] The offshore structure was built at this step.

Around March 6, 2025, Manus was released on an invitation-only basis, positioned as a general-purpose AI agent able to plan on its own, call tools, and deliver results. It grew quickly.[8]

Between April and May 2025, the company raised US$75 million in a round led by Benchmark, and a Benchmark partner joined the board. The roughly US$500 million valuation reported by the press was not confirmed by the company.[9]

In mid-2025, the company moved its headquarters and core team to Singapore, with additional offices in Tokyo and San Francisco, relocating its headquarters and most of its China-based employees to Singapore, according to media reports. Research and development had begun onshore, but the main entity and the team moved offshore at this point.[10]

By December 2025, the company said its annual recurring revenue had passed US$100 million, with 105 employees spread across Singapore, Tokyo, and San Francisco.[11]

On December 29, 2025, Manus announced that it was "joining Meta," saying that it would still operate from Singapore. The press estimated the deal at roughly US$2.5 billion, but the parties never disclosed a formal price.[12]

On April 27, 2026, China halted the transaction, using the phrase "foreign acquisition of the Manus project."[13]

With the facts laid out this far, one point of usage should be clarified first. What can be confirmed is the US$75 million round, the board seat, and the headcount. The "roughly US$500 million valuation" and the "US$2 billion to US$2.5 billion purchase price" are media figures, not formal numbers disclosed by the parties.[14]

On this basis, two points can be gathered into conclusions.

First, the regulator used the word "project," not "equity in a particular company." This wording may indicate that the regulator viewed the object of review more broadly than the shares of a single Singapore entity. The published decision, however, does not specify which entities, assets, technologies, data, personnel, or business operations it treated as components of the “Manus project.”[15]

Second, placing the timeline side by side, one trajectory is clear. The origin of the research and development, the core team, and the founders all came from onshore. Meanwhile the place of registration, the headquarters, and the operations on the surface had, by the time of the transaction, been moved to Singapore as a whole, and possibly connected to the Cayman Islands as well. In a word, this is a textbook "shell offshore, roots onshore" entity. Which country its "nationality" belongs to is precisely the point in dispute in this case.

II. China's Characterization: Why This Counts as "Foreign Investment"

Having seen the target clearly, we turn to how China brought it within its own jurisdiction. The difficulty is direct. On its face, the transaction was "an offshore party buying an offshore party" — an American company buying a Singaporean company. On what basis does it become "foreign investment" subject to Chinese jurisdiction?

To answer this fully, one must proceed step by step.

First, where the territorial anchor lies. The foothold for China's jurisdiction is not, in the first instance, that offshore equity transaction. It is the provenance of the Manus project itself. The timeline above already shows it: research and development began onshore, the core team came from onshore, Beijing Butterfly Effect was the early R&D entity, and as of early 2025 the founder still held 28% of it. The formation of the project, its early operation, and its subsequent migration to Singapore all took place, at least in substantial part, within China, or were carried out by onshore entities. This gives China a territorial connecting point. It is not an abstract "Chinese bloodline," but activity that actually occurred onshore.

Second, the point of the word "project." The regulator's phrasing was "foreign acquisition of the Manus project," not "acquisition of equity in a Singaporean company." That choice of words was not casual. It means the regulator did not confine the object of review to the single act of "Meta buying the equity of the Singapore entity." It treated this acquisition, together with the whole prior course of Manus's cross-border restructuring — the offshore entity taking over, the movement of personnel, the reorganization of the business — as one continuous matter. The legal shell of the enterprise can be cut into several pieces: a Chinese entity, a Singaporean entity, an offshore equity transaction. But as a technology project, Manus is continuous in economic and organizational terms. The entity changed and the border was crossed, yet it is almost still the same product, the same technology, the same team. What the regulator fixed on was precisely this functional continuity.

Third, how the statute connects the onshore anchor to the offshore transaction. An onshore connecting point alone is not enough. There must also be legal authority allowing the regulator to follow that connecting point and reach the offshore transaction. That authority lies in Article 2 of the Foreign Investment Law and in the Measures for the Security Review of Foreign Investment. Article 2 defines foreign investment as investment activity conducted within China by a foreign investor, "directly or indirectly." The words "or indirectly" are the very interface that brings the offshore transaction in.[16] The Measures likewise bring "indirect" investment activity within review, and expressly cover "acquiring the equity or assets of a domestic enterprise through merger or acquisition." Article 4(2) then locks onto the situation of investing in fields such as "important information technology and internet products and services" and "key technologies," where the investor "obtains actual control of the invested enterprise."[17] Put together, what the regulator did becomes clear. Borrowing the language of "indirect investment" plus "actual control," it re-characterized this seemingly purely offshore acquisition as "indirect foreign investment in an onshore project," and then opened a security review of that re-characterized object.

Fourth, "trigger" is not "basis." There is a point here that is easy to get tangled in. In the whole chain, Meta's offshore acquisition plays the role of a "trigger," not of the "jurisdictional basis." Its function is to make Manus's pre-existing cross-border arrangements produce, for the first time, a final consequence for control: the project as a whole was to land in the hands of an American giant. It was precisely this consequence that prompted the regulator to look back and re-examine the entire project. But this does not mean that China had first to prove it held direct jurisdiction over the Singapore equity transfer itself. The pivot on which China actually exercises jurisdiction is, throughout, the onshore activity and the onshore entities. The offshore transaction is only the fuse that brought the problem at that pivot to the surface.

Putting the steps above together, the jurisdictional structure in this case can be summarized as follows. An onshore activity establishes a territorial foothold; the operational continuity of the "Manus project" ties the onshore activity and the offshore acquisition into one; and the capacity to regulate onshore entities and onshore matters is then used to produce a real effect on that offshore transaction.

One more point is worth making explicit. In essence, this approach is not a modest territorial jurisdiction. It carries a clear "control-plus-effects" coloring. Where you are registered does not matter; so long as the ultimate destination of control would affect the onshore project and national security, China reaches in. This is like the U.S. CFIUS reach-through in TikTok, discussed later — only in the opposite direction.

Finally, a measure of restraint. The foregoing is a reconstruction of the jurisdictional logic. As far as the current public materials go, the only line that has been confirmed to apply, and that produced the prohibitory consequence, is the security review. It characterized this acquisition as foreign investment subject to review and halted it on that basis. As for exactly which other regulated acts occurred onshore, such as technology export or outbound investment, the regulator only listed them publicly as assessment directions and did not make a finding. This paper does not treat them as established.

III. The Laws That May Apply: Read in Three Tiers

In this case, China's authorities referred, at various points, to a string of laws. But those laws do not carry equal weight. Some have actually been applied. Some were merely listed publicly by the authorities as "possibly applicable," without any finding that they were in fact applied. Others appear only in media reports or external commentary. To lump them together would badly overstate the tools the regulator has actually used. The following sorts them into three tiers by the strength of the evidence.

One expression needs to be clarified first. People often speak of the Ministry of Commerce "naming" the deal. What this actually refers to is narrower. At its regular press conference on January 8, 2026, the Ministry stated publicly that it would work with the relevant departments to assess whether the acquisition was consistent with the laws and regulations on export control, technology import and export, outbound investment, cross-border data transfer, and cross-border M&A.[18] That was only a list of "assessment directions" announced by the authority. It was not a public finding that these rules had in fact been applied. Nor was it a conclusion of illegality. Accordingly, apart from the security review decision, every other norm should be understood as "potential range," not as "already unlawful."

For ease of comparison, the table below sets out the difference in force among the three tiers.

Tier one is the core basis that has actually been applied: the security review of foreign investment. This is the highest in force and the clearest in outcome among the available materials. Its parent statute is Article 59 of the National Security Law. The state conducts national security review of foreign investment, of specific items and key technologies, and of network information technology products and services that affect or may affect national security.[19] The specific procedure and consequences fall under the Measures for the Security Review of Foreign Investment. Article 9(2) provides that, where a special review finds an effect on national security, a decision prohibiting the investment shall be made. Article 12 provides that, where the investment has already been implemented, the equity or assets shall be disposed of within a set period and the state of affairs restored to what it was before the investment. Article 13 provides for supervision of implementation.[20] The April 27 decision in Manus followed this path.

Tier two comprises the norms that the authorities publicly listed as "possibly applicable" but that, to date, have not been found actually to apply. They should be understood as potential range.

The first is technology import and export. The basis is Article 18 of the Foreign Trade Law and the Regulations on the Administration of Technology Import and Export.[21] There is a trap here that is easy to fall into. The Catalogue of Technologies Prohibited or Restricted from Export does contain an entry related to information processing (item 96, code 086501X). But its control points are qualified. For technologies such as speech recognition, speech synthesis, and AI interaction interfaces, the control is limited to those "specifically designed for the Chinese language and minority-nationality languages."[22] One cannot say broadly that "all AI technology is controlled." Whether Manus's technology falls within it cannot be shown on the available public materials.

The second is cross-border data transfer. The basis is Article 24 (data security review) and Article 31 (outbound transfer of important data) of the Data Security Law, together with the provisions of the Regulations on the Administration of Network Data Security concerning the obligations of the data recipient and the handling of important data in a merger or restructuring (Articles 14, 32, and 37).[23] It becomes relevant only where "important data" is in fact transferred, or where onshore personal information is involved.

The third is outbound investment and cross-border M&A. If an onshore entity did in fact contribute capital, equity, technology, or intellectual property to the Singapore entity, this could trigger the outbound-investment procedures of the NDRC and the Ministry of Commerce. It could also trigger the approval requirement for "round-trip M&A" under Article 11 of the Provisions on the Merger and Acquisition of Domestic Enterprises by Foreign Investors.[24]

Tier three appears only in media reports or in a U.S. report and must be treated as downgraded: taxation and the cross-border movement of funds. The source for this tier is a February 2026 brief of the U.S.-China Economic and Security Review Commission, which relayed media reports that the investigation might extend to taxation and cross-border currency flows. This is not a basis publicly confirmed by China's authorities. It cannot be placed alongside the assessment directions that the Ministry of Commerce actually listed, as though it were an established fact.

IV. Several Observations on China's Approach

Having laid out the law, the following offers five observations on this approach, from the perspective of a Chinese disputes lawyer.

First, the "onshore project plus actual control" test has strong reach.

Its logic looks to economic substance, not to the place of registration. Even if a company moves its shell to Singapore or the Cayman Islands, it can be reached, so long as its research and development, its team, and its control remain connected to China. The advantage of this logic is that it is hard to evade through structuring. The cost is that its boundary is relatively blurred. How much connection is "enough" is hard to draw as a clear line in advance. For the parties to a transaction, that means uncertainty.

Second, although the decision cannot be appealed or reversed, the state rarely acts abruptly. The statutory framework provides opportunities for consultation and other forms of pre-decision engagement. Such engagement would ordinarily be expected before a prohibition decision is issued, although the published Manus decision does not disclose whether, or to what extent, it occurred in this case.

The April 27 decision was extremely spare in its text. It listed no reasons, described no procedure, and named no party.[25] And Article 35 of the Foreign Investment Law provides expressly that a security review decision made in accordance with law is final. It is not subject to administrative reconsideration, nor to administrative litigation.[26] Reading these two points alone, it is easy to conclude that the party has no room at all.

But that conclusion is incomplete. The security review regime itself builds in a stage of prior interaction. Article 5 of the Measures allows a party to consult before filing. Article 6(3) requires the party to submit "an explanation of whether the foreign investment affects national security." Article 10 provides that, during the review, the Office may request supplementary materials and ask questions, and the party shall cooperate. Article 11 allows the party to modify the investment plan or to withdraw the investment during the review. Article 9 provides an exit through "conditional clearance," so long as the attached conditions can eliminate the effect on national security and the party undertakes in writing to accept them.[27]

Read together, these provisions bring a picture into view that is closer to practice. When the state first takes an interest in a transaction, it does not usually issue a prohibition outright. It first communicates with the party, asks for explanations, and offers a chance to adjust or to accept conditions. It is precisely at this stage that the party engages, explains, and bargains with the state over "whether national security is affected." Reaching the point of a prohibition usually means one of two things. Either the party failed to persuade the regulator from a national-security standpoint, or the two sides could not agree on the attached conditions. In other words, that spare and final decision is likely the result of a failed round of engagement. It is not an ambush without warning. The lesson for practitioners follows directly. The center of gravity in responding lies in the communication and explanation before the decision, not in relief after it.

Third, the remedy of "unwind the transaction and restore the prior state" is simple to state and extremely hard to carry out.

The decision required the parties to withdraw the transaction and restore the state of affairs to what it was before the investment.[28] But "restoring the prior state" can almost never be done cleanly in a business like AI. Equity may be transferable back. But can data that has already been shared be completely deleted? And how does one reverse the integration of teams and systems that have already been closed and merged? Looking at what followed, Manus, in its August 11, 2026 letter to users, confirmed that it was separating from Meta and restoring independent operation. Part of the data generated during the transaction period would be deleted, and then backed up and restored by users.[29] This shows that the "unwinding" is indeed happening. But it also shows that it is a long and complex process that is hard to complete fully.

Fourth, halting this transaction is, in essence, also a form of protection for the domestic AI industry. And it serves as a warning to those who come later.

This point must be seen in the context of China's AI industry. Starting a business onshore has its own distinctive conditions: vast data, relatively inexpensive technical labor, convenient infrastructure, and cheap electricity. From this arises a worrying pattern. Some teams use these conditions to build a successful product onshore. They then no longer seek to keep cultivating it. Instead, they "change the shell" and move offshore, seeking to be acquired by a U.S. technology giant and to cash out. Manus began its research and development onshore, rapidly moved its main entity and team to Singapore, and was then acquired by Meta. It hit every node on this path.

At the same time, in recent years U.S. technology giants have drawn in enormous amounts of capital from across the Western world through the capital markets. Against this background, when such a giant moves on a high-growth offshore AI company, the move may be a "protective acquisition" that values the target. But it may equally be a "strangling acquisition" that removes a potential competitor. If transactions of this kind proceed unimpeded, the long-run consequence is serious. China's most promising AI achievements would be fed offshore, to foreign giants, at the very moment they take shape. That, in turn, would hold back the accumulation and progress of the country's own AI technology.

Seen from this angle, China's halting of the Manus transaction is not merely a discretionary call in a single case. It also sends a clear signal to those who come later. Achievements that begin onshore and are built on onshore resources cannot simply be sold off as a matter of course. This dimension of industrial protection and warning is a side that cannot be ignored in understanding the deeper motive behind the case.

Fifth, the object of regulation appears to extend from the "transaction" to the "person."

According to the Wall Street Journal, two executives were interviewed and advised not to leave the country for the time being.[30] But two points must be made clear. First, the report itself indicates that this was guidance, not a formal exit ban. Second, when the Ministry of Commerce responded on April 2, 2026, it said it was not aware of the situation.[31] This fact appears only in media reports. It has not been confirmed by any separate announcement from the regulator.

Part Two: Two Points of Reference — A Mirror and a Footnote

With the Manus thread complete, two offshore cases serve as points of reference. They let us see China's approach in fuller dimension. The two cases carry different weight. TikTok is a mirror. GE-Honeywell is a footnote.

I. TikTok and ByteDance: A Mirror That Reflects in Reverse

TikTok provides a useful comparison in the reverse regulatory direction: Manus involved China scrutinizing a U.S. company’s acquisition of a Chinese-origin technology project, whereas TikTok involved the United States scrutinizing a Chinese-controlled company’s acquisition and operation of a U.S. business. The jurisdictional nexus in TikTok was nevertheless substantially more explicit, including an acquired U.S. business, U.S. users, U.S. data, and continuing U.S. operations.

Consider the facts first. ByteDance Ltd., the ultimate parent, is incorporated in the Cayman Islands. It was founded by Zhang Yiming, a Chinese national, and its important operations are in China. In November 2017, a ByteDance subsidiary acquired Musical.ly of the United States. The merger of the two formed what is today TikTok's U.S. business.[32] Is this not, once again, a company with "roots in China, a shell offshore, and its business done abroad"? Only this time, the party that halted the deal and forced a divestiture was the United States. The direction is reversed, but the underlying pattern is the same. This reflects back on the "offshore registration, onshore roots" conclusion from Part One.

Now consider the reasons. What is striking is that the concerns on both the Chinese and the U.S. sides are almost the same set. The United States invoked four legal paths in turn. First, under Section 721 of the Defense Production Act, CFIUS retroactively reviewed the 2017 acquisition. In August 2020, it ordered ByteDance to divest the assets and user data of TikTok's U.S. business.[33] Second, under the International Emergency Economic Powers Act, an executive order was issued that sought to prohibit future transactions within the United States relating to ByteDance. It was later revoked in 2021.[34] Third, in 2022, the No TikTok on Government Devices Act banned the app within the scope of government devices.[35] Fourth, in 2024, the Protecting Americans from Foreign Adversary Controlled Applications Act forced a divestiture through a "divest-or-be-barred-from-the-market" mechanism.[36] The core risks emphasized repeatedly across these four paths come down to three: the acquisition of vast sensitive data by a foreign adversary; the possible manipulation of the recommendation algorithm and content; and the possibility that the company could be coerced by its home state. Although the Manus decision does not disclose the specific risks that drove the outcome, the concerns identified in comparable cases resemble categories that may also arise under China’s national-security framework.

The differences are best seen in two places: the mode of relief and the transparency of procedure.

On relief, China's method is "unwind and restore the prior state." The U.S. method is a "qualified divestiture." Through a U.S. joint-venture structure called USDS, control and operational ties are severed, and the business is re-grounded locally in the United States. In early 2026, the TikTok USDS joint venture was established on this basis. U.S. investors held a majority, and ByteDance's stake fell below 20%.[37] One approach steps backward; the other reorganizes in place. By comparison, "restoring the prior state" is far harder. This bears out the point about execution difficulty made earlier in this Part.

On procedure, the difference is even clearer. The reasons on the U.S. side are laid out on the table. Congress produced a dedicated committee report setting out the legislative rationale.[38] The case was litigated all the way to the Supreme Court, and the judgment spelled out the data and algorithm risks in detail. In January 2025, the Supreme Court unanimously upheld the challenged provisions. It noted in particular that TikTok has roughly 170 million U.S. users. The scale and linkability of the data make the risk a reasonable inference supported by substantial evidence.[39] By contrast, China's decision is extremely spare in its text, gives no reasons, and cannot be appealed. Set against each other, the features of China's approach stand out sharply: a clear outcome, undisclosed reasons, and little room to challenge. This is not a judgment about which is better. It is to say that, even where both sides stop a transaction, the room they leave the party to "know" and to "be heard" differs greatly.

II. GE and Honeywell: A Footnote That Makes the Point

As a footnote, GE-Honeywell needs only a light touch. But the thing it reveals is important.

This transaction took place in 2001. GE was a New York company. Honeywell was a Delaware company. The two merged through a stock swap, at 1.055 GE shares for each Honeywell share. It was a textbook merger between American companies. The parties completed their U.S. pre-merger filing in November 2000 and filed with the European Commission in February 2001. Yet this merger between American companies was, in the end, blocked by the European Commission.[40]

Three points are worth keeping in mind.

First, on what basis did the EU regulate two American companies? Not on "whether you are a foreign company," but on turnover thresholds plus effects on the common market, that is, the "effects doctrine." Where the parties' combined worldwide turnover exceeded EUR 5 billion, and each exceeded EUR 250 million within the Community, the transaction had a "Community dimension."[41] Nationality did not matter. If you affect my market, I can regulate you. The Gencor precedent had already confirmed that a transaction between non-EU companies can fall under the Merger Regulation where it has a foreseeable, immediate, and substantial effect on the Community.[42]

Second, even allies will disagree. At the time, the U.S. Department of Justice required divestiture only on limited issues, such as military helicopter engines, and considered that the deal could largely be cleared.[43] The European Commission, however, took a broader view, looking to the group's overall strength, bundled sales, and the like. It found that the transaction would create or strengthen a dominant position. On July 3, 2001, it prohibited the transaction under Article 8(3) of the Merger Regulation.[44] Two allies reached completely different conclusions on a pure question of competition law.

Third, one jurisdiction saying "no" was enough to bring down the entire global transaction. Because "obtaining EU approval" was itself a closing condition expressly stated in the merger agreement, once the EU blocked the deal the global transaction could not be completed. The parties formally terminated it on October 2, 2001.[45]

This footnote makes one point. Extending a regulatory hand across borders to intervene in another country's corporate transactions is neither new nor unique to the China-U.S. relationship. And it is not confined to the single field of national security. It is a structural phenomenon.

Part Three: The Limits of and Conflicts in Extraterritorial Jurisdiction

Having discussed the cases one by one, the final step lifts the analysis one level. What happens when these practices are stacked together?

First, the reasons states give are converging. Data, technology, susceptibility to coercion: China says this, and the United States says it too. The basket of "national security" grows larger and larger. In the end, states are borrowing one another's logic. You can use national security to stop me; I can use the same reason to stop you.

Second, once jurisdiction overlaps, transactions are easily deadlocked. A single cross-border transaction may be targeted at the same time by several states, each from a different angle, and each able to veto it on its own. This brings the risk of transactional deadlock, and it leaves room for mutual retaliation. GE demonstrated this long ago: a veto in one jurisdiction is enough to bring down a global transaction.

Third, coming down to the parties themselves, there are a few practical reminders. In building the transaction structure, the possible jurisdictional connecting points should be thought through in advance. Key regulatory approvals are best set as conditions precedent to closing, rather than closing ahead of them. How regulatory risk is allocated between buyer and seller should be spelled out in the contract.

Having come full circle, we return to the "shell offshore, roots onshore" enterprise. It thought that changing its place of registration would change its identity. But as long as the roots remain, it may not escape. This is Manus's situation. It is also the situation that many similar enterprises are about to face.

Conclusion

Return to Manus.

 Although the Manus decision neither discloses a complete jurisdictional test nor identifies the precise statutory bridge between the project’s onshore connections and foreign-investment review, the sequence of events and the decision’s reference to the “Manus project” support a plausible reconstruction of China’s approach. China appears to have treated Manus as a continuing technology project whose regulatory nationality was not necessarily altered by offshore incorporation and restructuring. On this reading, Meta’s acquisition was the triggering control event, the project’s formation and migration from China supplied the substantive connection, and China’s regulatory authority over the project’s onshore dimension enabled it to affect the offshore transaction. The Manus case therefore illustrates how a state may look beyond corporate form and use the continuing regulatory nationality of a technology project to establish and implement jurisdiction over a formally offshore acquisition.

It is worth noting that this answer is, at its core, quite similar to the approaches of the United States and the European Union. All of them look to substance and to effects. All of them are willing to extend a hand across borders. The real difference is not "whether to regulate." It lies in two places. After regulating, whether reasons are given and whether room is left to be heard. And after halting a deal, whether the demand is to restore the prior state or to permit reorganization in place.

For a company with a "shell offshore and roots onshore," Manus is a reminder. Changing the place of registration does not necessarily eliminate regulatory exposure where legally relevant onshore connections remain, and states are increasingly willing to scrutinize such connections in cross-border technology transactions.

[1] Office of the Working Mechanism for the Security Review of Foreign Investment, National Development and Reform Commission, Security Review Decision on the Foreign Acquisition of the Manus Project (Apr. 27, 2026), https://zfxxgk.ndrc.gov.cn/web/iteminfo.jsp?id=20623.

[2] Manus, Manus Joins Meta for Next Era of Innovation (Dec. 29, 2025), https://manus.im/en/blog/manus-joins-meta-for-next-era-of-innovation.

[3] Dan Primack, Meta's Deal for Manus AI Could Be Worth $2.5 Billion, Axios (Dec. 30, 2025), https://www.axios.com/2025/12/30/meta-manus-ai.

[4] Manus AI Launched in China, China Daily (Mar. 6, 2025), https://global.chinadaily.com.cn/a/202503/06/WS67c9ba2fa310c240449d9276.html.

[5] Manus's Leaders Can't Leave China, Wall St. J. (Mar. 26, 2026), https://www.wsj.com/public/resources/documents/QWgYHIzwtKDO4bQlV5jV-WSJNewsPaper-3-26-2026.pdf; see also China, AI Competition and the Manus–Meta Transaction, Wash. Post (Apr. 21, 2026), https://www.washingtonpost.com/world/2026/04/21/china-ai-competition-manus-meta/.

[6] Google Play, Manus AI — Developer: Butterfly Effect Pte. Ltd., https://play.google.com/store/apps/details?hl=en_CA&id=tech.butterfly.app; Butterfly Effect Pte. Ltd. (UEN 202330764R), SGPBusiness, https://www.sgpbusiness.com/company/Butterfly-Effect-Pte-Ltd.

[7] See China, AI Competition and the Manus–Meta Transaction, supra note 5.

[8] Manus AI Launched in China, supra note 4.

[9] Manus, Manus Update: $100M ARR, $125M Revenue Run-Rate (Dec. 17, 2025), https://manus.im/blog/manus-100m-arr (confirming the US$75 million round and the board seat); Chinese AI Startup Manus Scores Funding at $500 Million Value, Bloomberg (Apr. 25, 2025), https://news.bloomberglaw.com/artificial-intelligence/chinese-ai-startup-manus-scores-funding-at-500-million-value (the ~US$500 million valuation is a media figure).

[10] Manus's Leaders Can't Leave China, supra note 5; see also Manus Update, supra note 9 (105 employees across Singapore, Tokyo, and San Francisco).

[11] Manus Update, supra note 9.

[12] Manus Joins Meta, supra note 2; Primack, supra note 3.

[13] Security Review Decision, supra note 1.

[14] Manus Update, supra note 9; Primack, supra note 3.

[15] Security Review Decision, supra note 1.

[16] Foreign Investment Law of the People's Republic of China, art. 2 (Presidential Order No. 26, promulgated Mar. 15, 2019, effective Jan. 1, 2020).

[17] Measures for the Security Review of Foreign Investment, arts. 2, 4(2) (NDRC & MOFCOM Order No. 37, promulgated Dec. 19, 2020, effective Jan. 18, 2021).

[18] Ministry of Commerce of the People's Republic of China, Regular Press Conference (Jan. 8, 2026), https://www.mofcom.gov.cn/xwfbzt/2026/swbzklxxwfbh2026n1y8r/index.html.

[19] National Security Law of the People's Republic of China, art. 59 (Presidential Order No. 29, effective July 1, 2015); see also id. art. 60.

[20] Measures for the Security Review of Foreign Investment, arts. 9(2), 12, 13, supra note 17.

[21] Foreign Trade Law of the People's Republic of China, art. 18 (2025 Revision, effective Mar. 1, 2026); Regulations on the Administration of Technology Import and Export (State Council Order No. 331, as amended).

[22] Catalogue of Technologies Prohibited or Restricted from Export, item 96 (code 086501X) (MOFCOM & MOST Announcement No. 57 (2023), as partially adjusted by Announcement No. 28 (2025)).

[23] Data Security Law of the People's Republic of China, arts. 24, 31 (effective Sept. 1, 2021); Regulations on the Administration of Network Data Security, arts. 14, 32, 37 (State Council Order No. 790, effective Jan. 1, 2025).

[24] Measures for the Administration of Outbound Investment by Enterprises (NDRC Order No. 11, effective Mar. 1, 2018); Measures for Overseas Investment Management (MOFCOM Order No. 3, effective Oct. 6, 2014); Provisions on the Merger and Acquisition of Domestic Enterprises by Foreign Investors, art. 11 (as amended by MOFCOM Order No. 6 (2009)).

[25] Security Review Decision, supra note 1.

[26] Foreign Investment Law, art. 35, supra note 16.

[27] Measures for the Security Review of Foreign Investment, arts. 5, 6(3), 9, 10, 11, supra note 17.

[28] Security Review Decision, supra note 1.

[29] Manus, A Note to Our Users (Aug. 11, 2026), https://manus.im/zh-cn/blog/a-note-to-our-users.

[30] Manus's Leaders Can't Leave China, supra note 5. The report characterizes the arrangement as guidance rather than a formal exit ban.

[31] Ministry of Commerce Responds to Hot Topics Including Meta's Acquisition of Manus, Xinhua / MOFCOM (Apr. 2, 2026), https://cacs.mofcom.gov.cn/article/gnwjmdt/sb/zm/202604/187694.html.

[32] Regarding the Acquisition of Musical.ly by ByteDance Ltd., 85 Fed. Reg. 51297 (2020).

[33] Defense Production Act of 1950 § 721, 50 U.S.C. § 4565; see also Regarding the Acquisition of Musical.ly by ByteDance Ltd., supra note 32.

[34] International Emergency Economic Powers Act, 50 U.S.C. §§ 1701–1708; Exec. Order No. 13942, 85 Fed. Reg. 48637 (Aug. 6, 2020), revoked by Exec. Order No. 14034, 86 Fed. Reg. 31423 (June 9, 2021).

[35] No TikTok on Government Devices Act, Pub. L. No. 117-328, div. R (2022).

[36] Protecting Americans from Foreign Adversary Controlled Applications Act, Pub. L. No. 118-50, div. H, 15 U.S.C. § 9901 note (2024).

[37] See Exec. Order No. 14352 (applying the “qualified divestiture” determination under PAFACA § 2(c), (g)(6)); Silver Lake, “TikTok USDS Joint Venture LLC Established in Compliance with U.S. Regulatory Requirements,” Jan. 22, 2026, https://www.silverlake.com/tiktok-usds-joint-venture-llc-established-in-compliance-with-u-s-regulatory-requirements/ (announcing the establishment of the joint venture and identifying ByteDance’s retained 19.9% interest).

[38] H.R. Rep. No. 118-417 (2024).

[39] TikTok Inc. v. Garland, 604 U.S. ___ (2025); see also TikTok Inc. v. Garland, 122 F.4th 930 (D.C. Cir. 2024).

[40] Commission Decision of 3 July 2001, Case COMP/M.2220 — General Electric/Honeywell, 2004 O.J. (L 48) 1. GE notified the European Commission on February 5, 2001.

[41] Council Regulation (EEC) No 4064/89, art. 1 (on the control of concentrations between undertakings).

[42] Gencor Ltd v Commission, Case T-102/96, [1999] ECR II-753.

[43] U.S. Dep't of Justice, Justice Department Requires Divestitures in Merger Between General Electric and Honeywell (May 2, 2001).

[44] Council Regulation (EEC) No 4064/89, art. 8(3); Commission Decision, Case COMP/M.2220, supra note 40.

[45] Commission Decision, Case COMP/M.2220, supra note 40. The subsequent actions for annulment were dismissed in General Electric v Commission, Case T-210/01, and Honeywell v Commission, Case T-209/01 (Ct. First Instance, Dec. 14, 2005), which upheld the prohibition on three market-specific grounds while faulting other parts of the Commission's reasoning.