A ban order wipes out 2 billion USD

April 27, 2026, the Office of the Work Mechanism for Foreign Investment Security Review (National Development and Reform Commission) lawfully and regulatively made a decision to prohibit foreign investment in the Manus acquisition project, requiring the parties to revoke the transaction.

Just a few dozen words directly pressed the stop button on this over 2 billion USD deal. Years of product refinement, legal framework segmentation, financing and exit arrangements—all collapsed suddenly, gone with the wind.

This is the first publicly halted foreign acquisition in the AI field since the “Measures for the Security Review of Foreign Investment” came into effect in January 2021.

There is a special aspect to this deal:

Both parties in this transaction have legally become offshore: Meta is an American company, Manus has completed registration in Singapore, and established a holding structure in the Cayman Islands. Yet, Chinese regulators ultimately still made the decision to prohibit the investment.

The spillover effects of this case, along with the hidden sides—companies like Moon, ByteDance, LeapStar—are now facing clearer compliance guidance. Behind this, lies a deeper issue: traditional offshore structuring methods are now completely invalid.

Entrepreneurs must think through their compliance routes from Day 0.

This article is not about storytelling—it’s about the essentials: Which laws and regulations do regulators rely on? Where are the red lines in the “bathing” style overseas expansion? Starting today, how should enterprises choose?

Lawful and Regulative—What laws and regulations do we rely on?

Looking back at the Manus case, initial industry discussions mostly focused on “what happened”—migration, segmentation, bans. But as details emerged, legal attention shifted to a more fundamental question: On what basis can regulators halt this deal? What laws and regulations are they relying on?

The answer isn’t in a single law but a three-layered regulatory logic. These layers work together, forming an unavoidable review process.

The first layer: Recognizing “Chinese Subject”—the foundation of the piercing review

This is the legal starting point of the entire case: What exactly is Manus?

Legally, the answer seems clear—Manus has completed registration in Singapore, with a Cayman holding structure. Its parent company, Butterfly Effect Pte, is a bona fide Singapore entity. This was the core legal argument of the Manus team throughout the transaction:

“Our main structure has shifted to an offshore setup.”

But regulators’ response was:

Formality doesn’t count; substance does.

Jingtian & Gongcheng lawyers systematically analyzed why the “legal shell offshore” failed in the Manus case. The root lies in the fact that the core assets of AI are substantively connected to China across four dimensions:

• Team dimension: The engineers mastering core logic have long accumulated R&D experience domestically; their technical capabilities are trained and cultivated within China;

• Computing power dimension: R&D within China has formed technical interfaces and paths for resource scheduling; the architecture of the core system bears Chinese characteristics;

• Algorithm dimension: The development and training of core model weights were completed domestically, representing the most legally significant “technological source”;

• Data dimension: The training data from massive user interactions for human feedback reinforcement learning (RLHF) are highly concentrated within China.

All four dimensions point to the same conclusion: Manus’s legal form is Singaporean, but its “technological essence” as a company—its source, core, and foundation—are entirely within China. According to the principle of “substance over form,” from a regulatory perspective, these substantive connections are enough to justify piercing review—this is the first cornerstone for all subsequent legal actions.

Therefore, although in 2022 Xiao Hong founded Butterfly Effect Tech in Beijing, in 2023 established a “Cayman–Hong Kong–Beijing” red-chip structure, and in 2025 registered in Singapore with team separation and business isolation, legal recognition does not consider “when the exit occurred” but “where it originated.” Any technical assets originating in China do not change nationality just because of registration changes.

The second layer: Export restrictions and regulatory evasion—Legal characterization of “bathing” overseas expansion

Once the first layer confirms: Manus is substantively a “domestic enterprise,” the second layer’s legal logic follows: transferring core assets offshore is itself an export activity. Export activities are subject to export control regulations.

Manus’s three-step actions form a complete puzzle of “avoiding export controls” in regulators’ eyes:

Step 1, Subject transfer. Moving the company from China to Singapore, establishing the offshore entity Butterfly Effect Pte, and setting up a Cayman holding structure. Legally, this completes the first step of “de-Chinawhitening.”

Step 2, Team and asset migration. Rapidly dismissing nearly two-thirds of Chinese employees (80 out of 120), retaining over 40 core technical personnel to relocate to Singapore.

Step 3, Data and business segmentation. Clearing domestic social media accounts, blocking Chinese IP access, terminating local collaborations like Alibaba Tongyi Qianwen.

Legally, the transfer of core technical personnel carrying technical knowledge, R&D capabilities, and algorithm experience constitutes a potential “technological export” covered by the “Catalogue of Technologies Restricted for Export.” Additionally, according to the “Data Security Law” and “Measures for Security Assessment of Data Export,” the large volume of user interaction data accumulated before segmentation—originating mainly within China—has its data genes embedded in the model, making it impossible to delete or trace back after segmentation.

Thus, the regulator’s piercing logic can be summarized as:

Code written on Chinese land, data grown from Chinese users—this is “Chinese assets.” Transferring them is export, and export must be regulated.

The essence of “bathing” overseas expansion is to mask substantive violations with formal compliance—systematic evasion of export control regimes.

The third layer: Voluntary declaration mechanism—“I can’t say I don’t know”

If the first two layers are about “substantive violations,” the third layer concerns “procedural violations”—and is the easiest to convict.

Article 4 of the “Measures for the Security Review of Foreign Investment” explicitly states that foreign investments involving important information technology, key technologies, etc., “shall be voluntarily declared to the Office of the Work Mechanism before implementation.” This is a mandatory pre-investment declaration obligation—not a suggestion, nor a post-incident supplement.

Throughout the entire transaction process, Manus and Meta never voluntarily reported to Chinese regulators. During the months-long closing period, Manus and its investors seemed to have reached a dangerous tacit understanding: as long as regulators didn’t knock, they wouldn’t open the window.

In legal practice, “failure to report when required” is itself a serious violation. It signals either deliberate concealment or willful evasion. Either way, regulators will not overlook lightly.

A compliance lawyer summarized after the case:

“The biggest compliance flaw exposed by the Manus case isn’t the applicability of any particular regulation, but that the enterprise fundamentally abandoned its reporting obligations to Chinese regulators. In the legal system, evading procedures is even more intolerable than substantive violations.”

Looking back, Manus’s fate was sealed at the first layer: once piercing review identified it as a “substantive Chinese subject,” the export control logic of the second layer and the declaration obligations of the third layer automatically unlocked. These three layers form a logical closed loop—each link tightly connected, leaving no room for “luck.”

Why the National Development and Reform Commission?

The Ministry of Commerce was the first to act. On January 8, 2026, a spokesperson publicly stated that an assessment would be conducted on the compliance of this acquisition with export control, technology import/export, and foreign investment laws and regulations. But by April 27, the final decision was made by the NDRC.

There is an underlying story in this departmental switch. Some experts believe: the Ministry of Commerce relied on the “Catalogue of Technologies Restricted for Export,” which specifically describes controlled AI technologies for Chinese and minority languages. After the “bathing” process, Manus’s services had all shifted to English, excluding Chinese users. This could lead to disputes if only the export control route is pursued.

This is a space for legal applicability debate. But we lean toward a deeper implication: ultimately, legal applicability is subordinate to political considerations.

The NDRC oversees “security review,” while the Ministry of Commerce handles “technology import/export.” The NDRC’s intervention signifies that this matter has shifted from “business” to “sovereignty.”

In other words, as a macro department with broader economic management authority, the NDRC’s involvement sends a clear signal—not an isolated enforcement against a single company, but a systemic deterrent: “One punch to open the fight, to prevent a hundred punches.”

Killing one is to warn a hundred.

All industry players still observing now see where the red line is—not in vague legal clauses, but in the irrefutable ultimate measure of national security.

Four high-risk trigger points

Based on the Manus case and the “piercing review” principle established by the “Measures for the Security Review of Foreign Investment,” four red lines are now clear. Crossing any one of them means the “bathing” overseas expansion route is no longer viable.

Red Line 1: Founders holding Chinese passports without renouncing Chinese nationality

Manus founder Xiao Hong holds Chinese nationality. China’s export control law’s jurisdiction covers natural persons. This means the founder himself could become a regulatory focus; arrangements cannot be understood solely at the corporate level.

A harsher reality lies across the Pacific: in North American VC geopolitical risk assessments, Chinese founders face tightening financing environments. Leading Silicon Valley VCs like a16z, under geopolitical pressure, are sharply reducing investments in founders with Chinese passports. Manus’s Series B was led by Benchmark, but afterward, Benchmark faced strong backlash from U.S. politics, with several Republican senators calling the deal “assisting the Chinese government.” Investors from Founders Fund openly said:

“The founder is Chinese, the company is in Beijing, and the core technology is a general AI agent—that’s ‘original sin.’”

Both sides are closing doors. If you have a Chinese passport, U.S. capital is wary; if you have Chinese technology, regulators won’t let go. The gap is narrower than most imagine.

Red Line 2: Receiving state-owned funds

It’s not only “state funds directly investing” that count as state assets. Guidance funds from various levels of government, state-owned LPs, policy bank loans—all fall within “state asset infusion.” Also, those “office, computing power, talent subsidies”—even if the application process is cumbersome and considered minor—will be recorded in the books when audited later.

Red Line 3: First line of code written in China

The initial development location of core code, the completion of algorithm training, the storage location of technical documents—all these seemingly “purely technical” facts constitute proof of “technological source” in law. Manus’s early development was completed in China; when the team moved to Singapore, the code they carried already constituted a technological export. Yet, Manus never reported this transfer as a technological export.

Red Line 4: Using Chinese data

Many AI entrepreneurs have a misconception: as long as they clear Chinese user data and block Chinese IPs later, they are clean.

But regulators see “technological substance” not only in code but also in data genes.

The “Data Security Law” and “Measures for Security Assessment of Data Export” have clear review requirements for cross-border transfer of “important data.” Although Manus shut down Chinese services and blocked Chinese IPs, the user interaction data accumulated early on has already embedded in the model’s weights—data genes that cannot be deleted or traced back after the fact. Data grown from Chinese users means the model bears a Chinese label.

Entrepreneurs in specific industries: choose your side, starting now

The “Security Review Measures” set up a security review mechanism for foreign investments that may impact national security, focusing on fields like military and defense, as well as key areas of foreign control—such as important information technology, critical technologies, major infrastructure, and vital resources.

In the current regulatory environment after the Manus case, the following points deserve special attention:

First, the judgment of “actual control” in practice is not just about shareholding percentage; if foreign investors can significantly influence business decisions, personnel, finance, or technology (e.g., holding veto rights or key technical information), it falls into this scope. This broad definition means: even holding only 5% of USD funds but with veto rights attached could be deemed “significant influence,” thus “actual control,” triggering review.

Second, the NDRC, as the leading department of the work mechanism, has the authority to provide compliance guidance based on national security judgments. For example, on April 24, 2026, the NDRC instructed some AI companies to refuse U.S. capital—though not explicitly in the law, it falls under the “security review” scope authorized by Articles 3 and 7 of the Measures.

Third, it is not advisable to evade review via VIE, trust, or nominee arrangements. In practice, if such arrangements are deemed to be evasion, companies risk correction, suspension, withdrawal, or other compliance measures.

Conclusion: The gray path of “balancing on both sides” has been completely closed off from all directions.

From now on, companies must clarify their compliance stance from Day 0.

Route A: Go the U.S. route—completely divest

If you decide to use U.S. funds, follow Silicon Valley’s path, with the ultimate goal of acquisition or U.S. stock listing, your task is not “bathing” but “blood transfusion.”

A strict standard: do not cross any of the four red lines.

Specifically, four actions:

First, resolve the founder’s nationality. Holding a Chinese passport is a compliance risk label in U.S. VC eyes. If you are committed to this route, renouncing Chinese nationality is not optional but a prerequisite.

Second, avoid state-owned funds. Any funds involving government guidance funds, state-owned LPs, or policy bank loans must undergo thorough compliance due diligence early on, with necessary corrections or buybacks.

Third, the source of code must be offshore. This is the most brutal and core rule. The first line of core algorithm code must be written outside China. Domestic teams can only handle non-core modules or peripheral business. You need to establish a truly capable offshore R&D center from the start—not a shell, but an entity.

Fourth, isolate data and users from day one. Do not touch Chinese user data at all. Not “cleaning later,” but “never owned.”

The premise of this route is: you can bear the cost of a complete break with the domestic market. All revenue, users, and brand synergy in China are abandoned. You gamble that global returns will outweigh this cost. And even if you do all of the above, you still face an increasingly unfriendly U.S.—the founder’s Chinese identity remains an “original sin” in some Silicon Valley circles.

Route B: Go the domestic route—align with the national team

If you don’t want or can’t follow the U.S. path, then make compliance your moat.

Core logic: China’s land, China’s ground, can only grow China’s RMB.

First, actively embrace state-owned/private capital. Prioritize RMB funds, government guidance funds, and central enterprise investment platforms in financing. This is not a forced choice but a strategic binding: a state-owned background is the strongest pass for regulation.

Second, turn compliance into a first-mover advantage. While others try to circumvent, proactively declare security review, classify and grade data, and file technology exports. Regulators see you as “one of us”; the market sees your compliance investment as a barrier that others can’t quickly catch up with.

Third, turn qualification certifications into licensing barriers. Trust certifications, data security maturity, “specialized and innovative” recognitions—these are not costs but licenses. In a tightening regulatory environment, having licenses can be the difference between life and death.

Fourth, proactively declare for security review. According to Article 4 of the “Measures for the Security Review of Foreign Investment,” foreign investments involving important information technology and key technologies must be declared before implementation. For companies following the domestic route, this is not a burden but the best way to show your stance to regulators.

Choosing this path means accepting RMB fund valuation logic and exit rhythms—fast, high-volume acquisitions like $2 billion in a flash may be irrelevant, but you gain policy stability and continuous operation rights in the domestic market.

If you want to grow big, there is no third way left.

The “Cayman holding + Singapore operation + domestic R&D + USD financing” balancing model has already been sentenced to death. Hesitating on this path is not flexibility but danger. Regulators won’t give you exemptions just because you haven’t figured it out.

Choose the U.S. route—go clean. Choose the domestic route—go all in.

This is the only operational manual Manus leaves for all cross-border entrepreneurs.

Final Words: Butterfly Effect, a Prophetic Name

Manus named its parent company Butterfly Effect—蝴蝶效应. Looking back now, this name can only be seen as a self-fulfilling prophecy.

This butterfly flapped its wings twice, stirring two storms. One was the Silicon Valley acquisition invitation; the other was a ban issued by Beijing. Now, the front and back pressure from regulators has taken shape: the acquisition invitation has become a compliance mirage, and this case will be written into the financing memo of every future cross-border tech company.

Reviewing the perfect path of “9 months to exit, $2 billion acquisition,” it secretly harbored three layered minefields from the start:

• Tech minefield: The moment AI core code is generated in China, regulators are watching;

• Data minefield: Using Chinese data leaves no trace;

• Identity minefield: In this era, technology has nationality, and so do the people making it.

Lawful and regulative—principle in the past, an iron rule in the future.

Today’s focus isn’t on condemning anyone but on recognizing a trend: the gray space previously exploited by registration, structuring, and主体切换 is being continuously squeezed. For founders, going overseas is no longer a game of “dodging regulation first, then patching compliance,” but about planning your主体、资金、技术、数据和申报路径 from Day 0.

May every entrepreneurial team seeking a way out in this era’s cracks see clearly the rules, stand firm, and go further.