When China’s National Development and Reform Commission ordered Meta to unwind its $2 billion acquisition of AI startup Manus on 27 April, Western media reached predictably for its standard toolkit: “authoritarian overreach”, “arbitrary intervention”, an assault on the “democratisation of technology.” What this framing systematically obscures is the substantive legal, strategic and political logic behind the decision.
The two pieces collected here provide that missing context. Sara Vivacqua’s investigation, published by the progressive Brazilian outlet Diário do Centro do Mundo and translated into English by the author, dissects both the legal architecture of China’s decision and the character of the company it rejected. Manus is not simply a commercial product; it is an autonomous AI agent – capable of operating inside authenticated platforms, accessing local sessions and executing complex multi-step tasks – built by Chinese engineers in China, with Chinese state support, before a hasty relocation to Singapore (presumably for purposes of regulatory evasion). The NDRC’s ruling establishes a clear and consequential precedent: jurisdictional control follows where technology is built and who builds it, not where a holding company is incorporated.
But Sara goes further, placing the ruling in the context of what Meta actually is. The company’s integration into US military AI development, its Llama models deployed across federal agencies and Five Eyes intelligence partners, its partnership with defence contractor Anduril, and its documented history of global electoral interference through the Cambridge Analytica scandal – all of this reframes the acquisition not as a business deal but as a potential intelligence operation. The question the Western press refuses to ask is the obvious one: why would any sovereign state hand strategic AI infrastructure to a company that functions as an arm of the US national security apparatus?
Below Sara Vivacqua’s article, we reproduce a Global Times editorial making the complementary case from a Chinese regulatory perspective: that the decision is legally grounded, internationally consistent, and entirely compatible with China’s continued openness to foreign investment in non-sensitive sectors. The EU, the US and Japan all operate comparable review mechanisms; the difference is that when China uses them, it is treated in the Western media as evidence of authoritarianism rather than ordinary statecraft.
Together, these two pieces offer what the mainstream coverage has failed to provide: a clear-eyed account of a decision that is legally sound and strategically coherent.
How China Blocked Zuckerberg’s Espionage Project with Manus AI
Meta, owned by Mark Zuckerberg and the parent company of Facebook, Instagram and WhatsApp, found itself frustrated this Monday (27th April) in its attempt to appropriate Chinese artificial intelligence technology.
The National Development and Reform Commission (NDRC), China’s state planning body responsible for reviewing the security of foreign investments, has ordered the cancellation of the acquisition of AI startup Manus and compelled Meta to unwind the US$2 billion purchase for violating its regulations.
Manus is not a simple ChatGPT-like large language model (LLM) that delivers text responses. It is an autonomous AI agent — a system that receives a relatively broad task and can execute multiple sequential actions autonomously: researching, opening pages, extracting data, organising information, writing, creating files, generating code, testing, reviewing and delivering complete results.
Manus can even operate in your own local browser, with your logins, sessions and local IP. This allows it to execute tasks on authenticated platforms — CRMs, paid services or business systems that would normally block automation coming from external servers.
Manus was launched on 6 March 2025 by Butterfly Effect Pte Ltd and was presented as capable of surpassing the performance of tools such as OpenAI’s Deep Research.
The So-Called “Democratisation of Technology”
Initial versions of Manus were created by engineers at Beijing Butterfly Effect Technology, which founder Xiao Hong established in 2022, and many of the company’s key people, including Xiao, are Chinese citizens. In its development stages, Butterfly Effect received support from the Chinese government – including subsidised rent for its offices – and benefited from Wuhan’s large pool of engineers.
However, in July 2025, the startup sacked most of its Beijing team and moved its headquarters to Singapore, in what it described as “global expansion”.
It might be disputable or unclear whether this manoeuvre was intended to circumvent Chinese jurisdiction though “offshore washing”. However, what is clear is that the strategy failed. And indeed, where a company builds its product matters more to establish the jurisdiction than where its holding company is registered. In cases like Manus’s, developing technology in China before transferring assets to an offshore entity through restructuring is a red flag.
The decision against Meta is legally relevant and sets a precise legal precedent. Beijing has made it clear that a startup that developed technology on Chinese soil, with Chinese engineers, initial Chinese financing and state subsidies, does not escape Chinese jurisdiction simply because it formally reorganised in Singapore.
Much of the US and European press reported the blocking of the operation as an “arbitrary” or even “authoritarian” action by the Chinese government against the “democratisation of technology”. However, China has a clear and consistent state policy and a regulatory architecture that treats technology as a matter of digital sovereignty, data control, national security, cybersecurity, intellectual property, export control and foreign investment.
The legal basis is fragmented in different legislation, but it begins with the Chinese Foreign Investment Law, which adopts the “pre-establishment national treatment + negative list” system. The list provides that competent authorities must refuse licences, business registrations or other approvals when foreign investors attempt to invest in sectors covered by the negative list without complying with its conditions.
China is exercising the same kind of control over strategic technologies that the United States exercises through the Committee on Foreign Investment in the United States (CFIUS), which has blocked dozens of Chinese acquisitions on grounds of “national security”.
Meta and Digital Colonialism
But beyond the matter of technological sovereignty and legal application, there is the issue of Meta seeking to appropriate foreign technology. Meta is not just a digital media enterprise seeking profits — it never was. It has a long-documented history of political interference in elections in various countries around the world and of its intimate connection as a service provider to the US military and mass surveillance apparatus.
On 22 September 2025, the GSA (General Services Administration), the central purchasing and services agency of the US federal government, announced a OneGov initiative with Meta to facilitate government use of its Llama AI models across all federal departments and agencies.
Llama has been used to help develop advanced AI tools for the US armed forces and national security agencies. At the very least this directly contradicts Meta’s own terms of service, which prohibit Llama’s use for “warfare, nuclear operations, espionage, or human trafficking”. But more importantly, it invites the question of what is the nature of Meta’s and other private companies’ participation in the ICE’s racial cleansing policy and US wars? And, furthermore, when private tech companies become integral to state military and immigration enforcement, can they be held directly accountable for human rights violations under international law, the so-called doctrine of human rights horizontal effect? We should all be asking those questions.
Meta has also signed a military contract in partnership with Anduril — a US defence technology company owned by Oculus founder Palmer Luckey — to build extended reality (XR) devices designed to “provide combatants with enhanced perception”.
What is perhaps most serious: since the end of last year, Meta has also made Llama available for national security use cases to the US’s Five Eyes security partners — Australia, Canada, New Zealand and the United Kingdom — and their private sector partners.
Five Eyes is the major espionage and mass surveillance network of these five countries, with extensive unofficial intelligence cooperation with Israel. Data is shared between the signatories’ secret services.
Meta has even announced that it is expanding this access to various “key democratic US allies” in Europe and Asia: France, Germany, Italy, Japan and South Korea, as well as NATO and European Union institutions.
The Cambridge Analytica Scandal
We cannot forget the Cambridge Analytica/Steve Bannon scandal. Facebook illegally shared data from up to 87 million users with Cambridge Analytica. The “This Is Your Digital Life” app collected data from at least 30 million users, although only 270,000 people downloaded the app, through access to Facebook “friends'” data via the Open Graph platform.
This data was used not only for common political advertising, but to extract mass behavioural data, infer psychological traits and deliver personalised political messages to very specific segments of the electorate, interfering in elections in dozens of countries.
Following disclosure of the violation in the US, Brazilian prosecutors revealed they were opening an investigation into whether Cambridge Analytica also improperly collected data from millions of Brazilian internet users to be used in the 2018 presidential elections.
Cambridge Analytica’s director-general of political operations said in a video recorded by Channel 4 that “We did this in Mexico, we did this in Malaysia, we are now moving to Brazil, Australia, China”. The alliance between Bannon (who was then Cambridge Analytica’s vice president) and the Bolsonaros was never a secret, and one infers here the clear intention to interfere in the presidential elections that elected Bolsonaro.
On 4 January 2020, a disclosure of more than 100,000 documents showed how Cambridge Analytica worked in 68 countries, developing a global infrastructure designed to manipulate voters on an “industrial scale”. The disclosure of documents began on New Year’s Day from an anonymous Twitter account called @HindsightFiles, which published material about elections in Malaysia, Kenya and Brazil.
Election Interference
In 2018, investigations linked Cambridge Analytica to Trump’s 2016 presidential campaign, the Brexit referendum in the United Kingdom and, according to a British Channel 4 investigation and subsequently disclosed documents, to Bolsonaro’s election in Brazil.
A secret Channel 4 News investigation revealed how Cambridge Analytica secretly campaigned in elections around the world. The British Parliament published a report on disinformation and “fake news”, strongly critical of Facebook’s role and the inadequacy of the regulatory regime.
In other words, Meta operates as an intelligence arm of the American state, providing AI for military use, national security and NATO and Five Eyes allies. The company has a proven track record of global electoral interference.
Why then should the Chinese state hand over its sovereignty to Zuckerberg and Meta, as has been suggested in the mainstream media?
How should we view China’s decision to halt the Manus acquisition?: Global Times editorial
The widely watched case of US tech giant Meta acquiring Chinese-founded AI startup Manus has taken a new turn. On April 27, the National Development and Reform Commission announced Monday that the office of the foreign investment security review working mechanism has, in accordance with laws and regulations, issued a decision to prohibit foreign investment in the Manus acquisition project and has ordered the parties involved to revoke the acquisition deal. This decision is reasonable, lawful, and necessary. It reflects the Chinese government’s fulfillment of its regulatory responsibilities and aligns with the international norm of strengthening security reviews in strategic technology sectors.
In March 2025, the startup Butterfly Effect launched Manus. As the world’s first general AI agent, Manus became an overnight sensation with a demo video, and access codes were, at one point, resold for tens of thousands of yuan. As a result, Manus was briefly regarded by the industry as “the second DeepSeek.” Just a few months later, however, Manus relocated its headquarters to Singapore, drastically downsized its domestic team, retained only core technical staff, and completely ceased services and operations within China. In December last year, Meta announced with great fanfare that it would acquire Manus for approximately $2 billion, making it the third-largest acquisition in Meta’s history.
The deal had been controversial from the outset. The biggest concern was that Manus, a company built on Chinese engineers and infrastructure, appeared to be “cutting ties” with its Chinese roots after securing US investment. At the time, many in the industry suggested that this could be a case of regulatory evasion through a “Singapore washing” strategy. In addition, the deal raised questions about “acquihiring,” with critics arguing that the primary valuation rested on the team rather than the company itself. A US law firm noted that talent retention is a core clause in such transactions, suggesting the acquisition essentially functioned as a recruitment strategy.
Given that the deal involves China, the US, and AI, China’s decision has drawn significant attention and, in some cases, overinterpretation. Although the Chinese side has not yet disclosed the specific reasons for blocking the acquisition, three points are already clear.
First, China has the right to intervene in this acquisition in accordance with laws and regulations. Some argue that Manus currently has no team in China and no products for the domestic market, and that its early development merely took place in China; therefore, they label China’s move to halt the acquisition as an exercise of “long-arm jurisdiction.” So-called “long-arm jurisdiction” refers to the extension of a country’s domestic laws beyond its borders, often using those laws to impose sanctions on third parties. However, China’s decision regarding the Manus acquisition is not of this nature. Although Manus had become a “Singapore-based company” when Meta announced the acquisition, the key issue is not where the company is registered or where its team is currently based. Rather, it lies in the extent of its technological, talent, and data links with China, and whether the transaction could harm China’s industrial security and development interests. Manus’s early R&D was conducted in China and that its core data originated there. These factors mean that the movement of its personnel, technology, and data is inevitably tied to China’s interests. Under rules issued on security review of foreign investment, China’s catalog of technologies prohibited and restricted for export, as well as the revised Foreign Trade Law, the export of such technologies and related cross-border transfers and investments are subject to security review and assessment until an approval is obtained. China therefore has a solid and well-founded legal basis to exercise jurisdiction over this transaction.
Second, China’s move to halt the deal is consistent with international practice. Some Western media outlets claim that the decision is driven by geopolitical competition with the US and targets American companies. This line of argument has a strong “seeing others as you are yourself” implication. Globally, cross-border mergers and acquisitions involving AI, data, algorithms, key software, core teams, and sensitive technologies have never been treated as ordinary commercial transactions. In recent years, countries have broadly strengthened investment security reviews, covering not only M&A deals but also greenfield investments such as new projects, and extending to sectors such as biomedicine and equipment manufacturing. China’s regulatory approach is fully in line with international practice.
Third, halting this acquisition does not mean that China’s business environment is “tightening.” China continues to optimize its foreign investment environment, having reduced the negative list for foreign investment multiple times, and is consistently easing access for foreign investment in sectors such as services and high-end manufacturing. Prohibiting a single sensitive AI-related acquisition is not contradictory to encouraging foreign investment and business operations in China. On the contrary, clarifying safety boundaries can provide reassurance to compliant foreign investors and enhance long-term confidence. The more open the economy becomes, the greater the need for improved regulations, clearer boundaries, and more stable expectations. The EU, the US, and Japan all have similar mechanisms, yet this has not prevented them from becoming hot global investment destinations. As the world’s second-largest recipient of foreign investment, China’s legal safety reviews and expanded cooperation are complementary and mutually reinforcing; both are essential.
AI is at the core of a new round of global industrial transformation. Leveraging its vast market advantages, complete industrial chain support, abundant application scenarios, and improving policy framework, China’s AI industry has entered a phase of rapid development, with a sustained burst of innovative vitality, making it a fertile ground for global AI innovation. We hope that more technology and innovation enterprises, including Manus, can find their place in this blue ocean in China, develop confidently, grow larger and stronger, and achieve better development and breakthroughs.
