In December 2025, Meta bought an AI company called Manus for $2 billion. Manus builds software that can carry out multi-step tasks on its own: booking, researching, writing reports, running workflows, the kind of work usually done by a junior employee. It was seen as one of the more promising Chinese AI companies of the past two years.
By April 2026, China's government told Meta to give it back.
The order came from China's National Development and Reform Commission, the powerful state body that controls major economic decisions. The official reason was national security: the regulator said the deal violated rules on foreign investment and technology exports. What made this order genuinely unusual is that it came five months after the deal had already closed. Meta had already integrated Manus's people and tools into its own systems. Forcing a reversal at that stage had never been done before under Chinese law.
Manus had tried to avoid exactly this outcome. The founders relocated the company's headquarters from China to Singapore in mid-2025, cut their China-based staff from around 120 to about 40, and set up a new Singapore legal entity to run the product outside China. The idea was to present Manus as a Singaporean company, not a Chinese one, and therefore beyond Beijing's reach. It did not work. Regulators concluded that the core technology, the data, and the intellectual origins of the company were Chinese, and that a new address does not change that.
The broader signal here is what the industry has started calling "Singapore washing," where Chinese founders relocate to Singapore to access Western capital while staying connected to Chinese talent and infrastructure. Beijing has now drawn a clear line: that route is closed.
The fallout is spreading well beyond this one deal. China's regulators have issued instructions to major Chinese AI companies, including ByteDance, Moonshot AI, and StepFun, telling them to reject American investment unless the government explicitly approves it. Researchers and executives at private AI firms now need government permission before leaving the country. These are not small procedural changes. They represent a structural tightening of who can invest in Chinese AI and who can leave to work elsewhere.
For anyone who sources technology vendors, evaluates software suppliers, or considers technology partnerships, this matters in a specific way. Any AI tool or service built by a team with Chinese roots, even one incorporated in Singapore or the UK or the US, now carries a different kind of regulatory risk. That company may not be able to freely move its people, share its data across borders, or accept outside capital without state approval. Due diligence on a vendor now has to include a question it rarely asked before: where were the founders originally based, and where was the core technology actually developed?
The Manus founders are now trying to raise $1 billion to buy the company back from Meta and restructure it as a Chinese joint venture, with a possible listing in Hong Kong. Those discussions are still early. The practical difficulties are real: several early investors have already been paid their proceeds from the Meta deal, Manus staff have moved into Meta's Singapore offices, and the technology was already being woven into Meta's advertising systems. Untangling all of that cleanly is genuinely hard.
What is not hard to read is the direction of travel. Both the United States and China now treat AI as a strategic asset in the same category as military hardware. The US restricts which chips China can buy. China restricts which companies Americans can invest in and which founders can leave the country. The space for neutral, purely commercial AI partnerships between the two countries is shrinking, and it is shrinking faster than most business operators have noticed.