China warns AI startups seeking to emulate Meta deal: Not so fast
Beijing worries the $2.5 billion acquisition of startup Manus will encourage more entrepreneurs to follow in its footsteps.
China said it would review Meta’s recent acquisition of artificial-intelligence startup Manus, firing a warning shot toward Chinese entrepreneurs lured by Silicon Valley riches.
People familiar with the matter said the review was part of broader consideration by Beijing about how it can build fences around its AI know-how when the U.S. and China are locked in an intensifying technology race. Officials have started identifying standout companies and reviewing AI-related technologies developed by Chinese researchers, with the aim of adding those of strategic importance to an export-control list, the people said.
Manus’s parent company, Butterfly Effect, was set up by Chinese co-founders in China, but last year it moved its headquarters to Singapore, seeking to expand its global user base. Last month, Meta said it would buy Manus in a deal that people involved said was valued at $2.5 billion.
A spokesman for China’s Ministry of Commerce said Thursday that Beijing was reviewing the deal. Cross-border acquisitions and the export of technology and data “must comply with Chinese laws and regulations," said spokesman He Yadong.
The Meta deal has prompted Beijing to worry that more Chinese startups, eager for growth and capital, will follow suit and pursue a U.S. buyout. Manus is the first high-profile AI startup with Chinese ties acquired by an American tech giant.
For Chinese startups, deals with top Silicon Valley players offer the potential for wealth far beyond what they can expect in China and a shot at capturing global renown. Silicon Valley players, for their part, say Chinese AI engineers are among the most talented in the world, particularly when it comes to developing consumer applications. A big part of Meta’s deal for Manus involves retaining most of its current employees, including Manus co-founders Xiao Hong and Ji Yichao, who are known as “Red" and “Peak," respectively.
Manus’s AI agent helps users write in-depth research reports, plan trips and prepare presentations—more involved and complex tasks than those typically given to chatbots such as ChatGPT or Gemini. Manus uses models developed by Anthropic and others, breaking up the overall assignment into smaller parts and drawing on the strengths of each model.
Despite Thursday’s announcement, Beijing appears to have few tools to block or influence the Meta deal. Manus’s founders are based in Singapore, and Meta has said the startup was discontinuing its services and operations in China following the deal. Meta and Manus didn’t respond to requests for comment.
Beijing’s initial review has focused on Manus’s agentic AI technology, people familiar with the assessment said. The review is in an early stage, and regulators may choose not to intervene, the people said. Industry participants said Manus’s technology didn’t appear to hold the same strategic importance as semiconductors, batteries and rare earths, and could eventually be matched by others.
Nonetheless, the opening of a review sends a message that Beijing intends to tighten its grip on the country’s AI know-how and top talent.
Some officials were irritated by Manus’s moves to relocate outside China and then sell itself to an American company because they were concerned that it showed a new funding path for other Chinese startups, The Wall Street Journal has reported. Manus has said its move to Singapore was out of business considerations because its product targets users outside China and many models it uses aren’t available in China.
There is precedent for Beijing restricting the export of AI technology.
China imposed export controls on AI-driven content-recommendation algorithms in 2020, when Chinese tech firm ByteDance was pressured by the White House to sell the U.S. operations of its popular short-video app TikTok. The algorithms are considered the secret sauce behind the app’s success and were built by ByteDance’s Chinese researchers.
Under the new rule, ByteDance had to seek approval from Beijing before it could sell TikTok, because a sale would involve transferring the technology to a foreign entity. That gave Beijing veto power in negotiations with Washington and leverage in trade talks. Following a September trade meeting between American and Chinese officials, TikTok last month said it had signed a deal to form a new U.S. joint venture.
Late last year, Chinese authorities considered imposing export restrictions on some technologies related to AI foundation models but met resistance from some officials and companies who argued such a move could hinder adoption of Chinese technology, people familiar with the discussions said. Chinese companies such as DeepSeek and Alibaba have released open-source AI models that are free to use, gaining a wide global following.
In recent years, facing intense competition at home, more Chinese companies have expanded their footprint globally, including by setting up subsidiaries in regions such as Southeast Asia and the Middle East. Many have played down their Chinese ties, hoping to reduce overseas scrutiny and geopolitical risks and get more access to global capital.
It doesn’t always go smoothly. Singapore-headquartered fast-fashion company Shein, which was founded in China, was required to seek approval from Chinese securities authorities for its initial public offering overseas because most of its suppliers and many employees are in China.
Lawyers who advise Chinese companies on such overseas moves say cases such as Shein’s and TikTok’s have made global investors wary because any unexpected regulatory hurdle could cause hefty financial costs.
Write to Yoko Kubota at yoko.kubota@wsj.com and Raffaele Huang at raffaele.huang@wsj.com

