For years, Chinese AI startups operated under an unspoken agreement with the market: build your technology in China, then move your holding company offshore when the time was right. That window is rapidly closing. In 2026, Beijing tightens its grip on AI firms that attempt to sever their Chinese roots using regulatory reviews, export control probes, and personal exit bans to keep both talent and intellectual property from crossing borders. The Manus case, involving Meta’s multi-billion-dollar acquisition of a Chinese-founded AI agent company, has become the defining test of where those limits now sit.
What distinguishes this moment from earlier rounds of Chinese tech regulation is the explicit reach beyond China’s borders. Previous crackdowns targeted companies operating on Chinese soil. This one is different: authorities are scrutinizing restructurings that were completed months or years earlier, signaling that geographic distance alone no longer offers protection. For the global AI investment community, the implications are substantial and still unfolding.
Why this matters: The Manus probe marks the first time Chinese authorities have used exit bans to target executives involved in a deal with a major US technology company a precedent that reshapes risk calculations for every founder and fund operating in this space.
What you need to know
Beijing tightens its grip on AI firms not just inside China, but on companies that have already relocated abroad.
The Manus-Meta deal triggered China’s first use of exit bans against executives involved in a major US tech acquisition.
“Singapore washing” registering offshore to escape scrutiny no longer works; regulators examine where the technology was actually built.
MiroMind and other labs received direct regulatory warnings, signaling the crackdown extends well beyond a single case.
Investors and founders are rapidly reassessing offshore restructuring strategies that were standard practice just 18 months ago.
The Timeline: How the Crackdown Unfolded
- Mid-2025
- Manus relocates to Singapore. Following a major funding round, the Chinese AI startup Butterfly Effect — creator of the Manus general-purpose AI agent moves its headquarters to Singapore, winds down Chinese operations, and replaces state-linked investors with US venture capital.
- Late 2025
- Meta acquires Manus for $2B+. The deal attracts immediate attention from Chinese regulators, who begin examining whether core intellectual property was transferred without required government approvals.
- Early 2026
- MOFCOM opens a formal probe. China’s Ministry of Commerce announces an investigation into potential violations of export control laws and technology transfer regulations specifically whether Manus moved protected AI intellectual property offshore without government sign-off.
- Late March 2026
- Exit bans imposed on Manus founders. CEO Xiao Hong and Chief Scientist Ji Yichao are summoned to Beijing by the National Development and Reform Commission and barred from international travel following questioning.
- April 2026
- MiroMind warnings issued; industry reckoning begins. Chinese regulators warn MiroMind an AI reasoning lab backed by Shanda Group — against relocating talent and research abroad, confirming the crackdown is sector-wide.
The Manus Case That Changed the Rules
Background
Manus began as the flagship product of Butterfly Effect, a Chinese AI startup that built a reputation for general-purpose AI agents capable of handling complex tasks — coding, market research, multi-step data analysis with minimal human input. It caught Silicon Valley’s attention early. By mid-2025, after a significant funding round, the company moved its headquarters to Singapore, replaced its Chinese institutional investors with US venture capital, and presented itself to the market as a Singaporean AI company. Meta acquired it for over two billion dollars shortly after.
The regulatory response
China’s Ministry of Commerce announced it would investigate the transaction for potential violations of export control laws and technology transfer rules. The central question: had Manus transferred core AI intellectual property to its Singapore entity without first obtaining the approvals required under the Regulations on Technology Import and Export Administration? Those rules mandate government sign-off before certain categories of technology developed in China can be moved abroad and advanced AI agents, authorities signaled, fall squarely within that classification.
“The threat of personal restriction gives Beijing significant leverage in any high-profile acquisition, even without blocking the deal outright.”
Technology policy analyst, cited in industry briefing, March 2026
Exit bans: a new kind of pressure
The investigation escalated sharply in late March 2026 when Chinese authorities imposed exit bans on two Manus co-founders: CEO Xiao Hong and Chief Scientist Ji Yichao. Both were summoned to Beijing for questioning by the National Development and Reform Commission and told they could not leave the country following those sessions. They retained freedom of movement within China but were barred from international travel. Meta has stated the transaction complied fully with applicable law and expressed confidence in reaching a resolution.
This marked the first time Beijing used exit restrictions to directly target executives involved in a deal with a major US technology company — a distinction that sent an unmistakable signal across the industry. Founders who had viewed offshore restructuring as routine risk management are now recalculating that assumption.
For deeper context on how this fits into the broader US-China technology rivalry, analysts and investors have been tracking the situation closely. You can visit here for the BBC’s ongoing coverage of China’s technology regulation and its global impact.
What “Singapore Washing” Is and Why It No Longer Works
The strategy Manus employed had a well-established name among Chinese founders: Singapore washing. The model was straightforward register a holding company in Singapore, hire a small local team, and present the firm to foreign investors and regulators as a Singaporean entity. It offered a practical path to international capital while maintaining the engineering talent base built in China. For a generation of AI entrepreneurs navigating restrictions from both Washington and Beijing, it was considered standard operating procedure.
Beijing’s intervention in the Manus deal has shattered the assumption that a Singapore address provides adequate cover. Advisers familiar with the process have begun warning founders that genuine non-Chinese status requires the entire team to relocate, the customer base to transition, and early Chinese investors to exit their positions not just a legal entity on paper. Where the product is actually built matters far more than where the holding company is registered.
What advisers now recommend: Founders seeking genuine offshore status must relocate engineering teams, transition customer relationships, clear Chinese institutional investors, and restructure IP ownership ideally before any significant technology has been formally recorded under Chinese jurisdiction. This process now takes 18–24 months minimum to credibly establish.
MiroMind and the Sector-Wide Warning
Manus was not an isolated case. MiroMind, an AI research lab focused on deep reasoning systems and backed by Shanda Group, received direct warnings from Chinese regulators against transferring valuable talent and research out of the country. The lab had been quietly relocating operations to Singapore. Its lead scientist who had previously worked at SenseTime departed the company after the firm instructed researchers to move abroad.
As of April 2026, MiroMind had no remaining staff in China and was actively raising funds in the United States, though most of its employees were Chinese nationals conducting work in Chinese. The company renamed one of its core products, a move engineers interpreted as part of a broader effort to distance the brand from its origins. Shanda Group’s founder was separately cautioned by Chinese regulators, though no formal investigation had been announced. The MiroMind situation confirmed what many in the industry had suspected: Beijing’s campaign extended well beyond any single high-profile acquisition.
China’s Regulatory Architecture Behind the Crackdown
The legal framework Beijing is deploying has been expanding for several years, often below the threshold of Western attention. The updated Cybersecurity Law, revised for 2026, now explicitly supports AI development while reinforcing state oversight over underlying infrastructure. The Data Security Law and the Personal Information Protection Law operate alongside it, creating a layered structure covering algorithms, data governance, and technology exports.
The most directly relevant instrument in cases like Manus is the Regulations on Technology Import and Export Administration, which require government approval before technology developed in China can be transferred to foreign entities. Advanced AI systems have been brought within their scope. China has not enacted a single comprehensive AI law that plan was removed from the 2025 legislative agenda in favor of targeted measures and pilot programs but the absence of one overarching statute has not weakened enforcement. Existing rules have proven sufficient to support the kind of review authorities are now conducting.
What This Means for Founders and Investors
For AI founders
The consequences for Chinese AI founders are significant and not entirely predictable. Some companies will respond by keeping operations inside China, making themselves less attractive to Western capital but avoiding regulatory exposure. Others may attempt to relocate at a much earlier stage before significant IP has been formally recorded under Chinese jurisdiction, where transfer rules have less purchase. A third group may abandon cross-border ambitions altogether. None of these paths is without cost.
For venture capital
Investors watching from Singapore and Silicon Valley are reassessing core assumptions. The belief that offshore restructuring offered a clean compliance pathway to US capital has been directly challenged. Venture capital firms that backed companies on the assumption that a Singapore entity would be treated as a Singapore entity are now examining the underlying details of how those companies were actually built. The Manus episode has created what some in the industry describe as a reckoning particularly for funds that quietly embraced the China-shedding model as a routine compliance solution.
The hardware dimension
The chip question remains entangled with this dynamic. Chinese authorities separately instructed domestic tech firms to avoid purchasing Nvidia’s H200 chips unless absolutely necessary, even as new US regulations created a limited export pathway. Analysts at the Council on Foreign Relations noted that the most likely purchasers of those chips companies including Alibaba and Byte Dance have documented ties to Chinese government and military institutions, complicating any straightforward commercial reading of the arrangement. The AI competition between the two countries is being contested simultaneously at the level of hardware access, talent retention, and regulatory architecture.
The Bigger Picture: A Mirrored Playbook
For most of the past decade, the United States was the primary actor imposing technology restrictions from semiconductor export controls to outbound investment screening. China was largely on the receiving end. The Manus probe and the MiroMind warnings represent a meaningful posture shift: Beijing is now willing to deploy an analogous playbook by making it harder for Chinese AI assets people, code, and data to flow toward American companies.
This is not merely regulatory defensiveness. It reflects a strategic judgment that AI capability is a national asset and that allowing it to be acquired by foreign entities even through nominally offshore structures constitutes a form of technology transfer that Beijing is no longer willing to permit without oversight and leverage.
Frequently Asked Questions
Why is Beijing tightening its grip on AI firms in 2026?
Beijing tightens its grip on AI firms because it classifies artificial intelligence as a strategic national asset one it is unwilling to see acquired by foreign companies through offshore corporate restructuring.
What is Singapore washing and why did it stop working?
Singapore washing refers to the practice of registering a holding company in Singapore and sometimes hiring a small local team so that a Chinese-founded startup could present itself as a Singaporean entity to access foreign investment and avoid scrutiny from US and Chinese regulators.
What happened to the Manus founders after the Meta acquisition?
Following China’s Ministry of Commerce probe into the Meta-Manus deal, co-founders CEO Xiao Hong and Chief Scientist Ji Yichao were summoned to Beijing for questioning by the National Development and Reform Commission.
How does China’s regulatory framework control AI technology exports?
China controls AI technology exports primarily through the Regulations on Technology Import and Export Administration, which require government approval before technology developed in China can be transferred to foreign entities.
What does Beijing’s AI crackdown mean for global investors?
For global investors particularly venture capital funds that backed companies on the assumption that a Singapore entity would be treated as a Singapore entity the Manus case is forcing a fundamental reassessment.