Opinion: The Limits of Singapore-Washing in AI

Opinion: The Manus Saga Shows The Real Limits Of "Singapore Washing" In Asian AI

Singapore has spent two years promoting itself as the neutral ground for Asian AI, the place where Chinese, American, and Indian companies can build without fully picking a side. The detention of Manus co-founders Xiao Hong and Ji Yichao on March 25, after a meeting with the National Development and Reform Commission in Beijing, made it plain that Beijing has rules for what it calls cross-border operations and that Chinese-founder companies do not get to relocate their decision-making by changing their letterhead. Singapore's neutral-ground pitch is real, but it is narrower than the marketing suggests.

What Actually Happened

Manus, the agentic-AI startup that briefly became the most-discussed Chinese consumer AI product of 2025, had restructured into a Singapore entity and was in advanced acquisition talks with Meta at a reported $2 billion valuation. On March 25 the founders were barred from leaving China after a meeting with the NDRC, and the Ministry of Commerce then opened a national security review of the deal. By April 2 Beijing's position had crystallised, in coverage by Asia Times, into an explicit message: cross-border operations are fine, but only those that "comply with Chinese laws and regulations and follow due procedure."

The message inside that message is what matters. Singapore-domiciled subsidiaries of Chinese companies are not, in Beijing's view, fully outside Chinese jurisdiction. The location of the corporate entity is a secondary consideration. The location of the founders, the engineering team, the training data, and the original IP all matter, and where any of those sit inside mainland China the relevant regulatory bodies will assert oversight regardless of where the holding company is registered.

Singapore offers a credible regulatory neutral ground. It does not, and cannot, offer a way out of Chinese jurisdiction for companies whose technical and human core remains in China.

Why The Singapore Pitch Still Works, Just Differently

The Manus episode does not invalidate Singapore's neutral-ground proposition. Plenty of Asian AI activity legitimately benefits from Singapore's regulatory environment, the Monetary Authority of Singapore supervisory framework, the IMDA sandboxes, and the city-state's English-language commercial law. DeepSeek, Alibaba, Tencent, and many smaller Chinese companies have established meaningful Singapore operations and use them as launchpads into Southeast Asia and the Middle East.

The narrower truth is that Singapore offers regulatory access to Western markets, neutral procurement positioning, and operational flexibility. It does not offer ownership or jurisdictional escape. A Chinese-founded AI company that wants to be sold to a US buyer will still need Beijing's blessing. A Chinese-trained foundation model whose weights are exported through a Singapore subsidiary still triggers Chinese export-control oversight. A Chinese-language dataset acquired by a Singapore entity still falls under Chinese data-security law.

By The Numbers

What Western Buyers And Asian Founders Should Take Away

For Western buyers the lesson is procedural. Acquiring a Singapore-domiciled company with Chinese-founder origin should now include explicit Chinese regulatory due diligence as standard, with a realistic timeline of three to six months for any necessary national security review. The cost of skipping that step is not just regulatory delay, it is reputational risk if the founders or technology become caught in an unexpected Chinese review process mid-deal.

For Asian founders the lesson is structural. A Singapore subsidiary alone does not provide jurisdictional flexibility. Founders who genuinely want a multi-jurisdictional posture need to build engineering teams, acquire training data, and structure IP ownership in a way that distributes risk across jurisdictions from day one. Retroactive restructuring after a deal is in motion is too late.

For Singapore policymakers the lesson is opportunity. The city-state's neutral-ground positioning is more credible, not less, when it is honest about its limits. The IMDA's ISO/IEC 42119-8 testing standard submission is exactly the kind of work that builds genuine neutral-ground capability. The marketing narrative around Singapore as a clean Chinese-AI exit ramp is a different proposition, and the Manus episode just made it harder to sell.

Singapore's value to Asian AI is not as a relocation strategy. It is as a place where rules are written, sandboxes work, and procurement is rational. That is durable. Relocation theatre is not.

The Wider Asian Picture Is Becoming More Honest

The Manus saga is happening at the same moment China is doubling down on sovereign-stack messaging. Huawei Ascend silicon now powers DeepSeek V4 training. The amended Cybersecurity Law has tightened obligations for cross-border data transfers. The Cyberspace Administration of China is more visibly involved in foundation-model approvals. Together these developments mean Beijing is being clearer about what is and is not acceptable, which paradoxically makes operating decisions easier for everyone.

For Indian, Japanese, Korean, and Southeast Asian founders the implication is a wider strategic opportunity. Companies that can credibly claim non-Chinese origin and non-Chinese supply chain now have a procurement advantage in markets that are increasingly cautious about Chinese AI provenance. Sarvam AI's $350 million round is partly a story about that opportunity. So is the Anthropic-NEC partnership in Japan and the broader Sea Limited consolidation in Singapore.

For broader regional context, see our coverage of China's amended Cybersecurity Law and the Stanford AI Index findings on Asia AI optimism and governance.