China has blocked Meta’s $2 billion acquisition of Manus, forcing the U.S. tech giant to walk away from a deal that had already been completed months earlier.
Summary
China ordered Meta to scrap its $2B Manus deal over national security concerns tied to AI assets.
Manus executives were reportedly restricted from leaving China during the regulatory review.
Authorities are tightening control over cross-border transfers of advanced AI technology and talent.
The cancellation order was issued by China’s National Development and Reform Commission, reflecting growing caution over transferring key artificial intelligence capabilities to foreign firms amid rising tensions with the United States.
Meta, the California-based parent of Facebook, had acquired Manus in December for over $2 billion to strengthen its push into AI agents—systems designed to perform complex digital tasks with minimal human input.
Regulatory pressure intensified in March, when Manus CEO Xiao Hong and chief scientist Ji Yichao were reportedly prevented from leaving China during the review process.
AI becomes a strategic battleground
China’s move underscores how artificial intelligence is increasingly central to geopolitical competition. Experts say Beijing is now treating AI assets similarly to semiconductors—sectors where outbound transfers are tightly controlled.
Despite Manus relocating its headquarters to Singapore, it remained under Chinese regulatory oversight, signaling that geographic shifts may not shield companies from intervention. The startup had been positioned as a potential rival to DeepSeek after unveiling a general-purpose AI agent capable of handling diverse tasks.
The decision could also heighten tensions ahead of a planned meeting between Donald Trump and Xi Jinping, where technology and trade issues are expected to dominate discussions.
Meta pushes ahead with AI ambitions
Despite the setback, Meta continues to expand its AI efforts internally. The company is investing in systems designed to automate workflows, with agents capable of performing tasks such as coding and digital navigation.
Under initiatives like the Model Ability program, Meta is also gathering real-world usage data to train its AI systems. According to CTO Andrew Bosworth, the goal is to build tools where AI agents handle most tasks, while humans guide and refine their performance.
Additionally, Meta is exploring new integrations across platforms like Instagram and Reels, aiming to create a more automated ecosystem that blends content discovery with commerce.
