China Orders Meta to Unwind $2 Billion Acquisition of AI Startup Manus Over National Security

China's National Development and Reform Commission (NDRC) has formally ordered Meta Platforms to unwind its $2 billion acquisition of Beijing-based AI startup Manus. The directive, issued early Saturday, mandates that the transaction be fully reversed within 90 days. This marks the first time Beijing has retroactively dismantled a completed acquisition of a domestic artificial intelligence firm by a US technology giant, sending immediate shockwaves through global tech markets.[1][3]

The NDRC's ruling explicitly relies on the "Protection of Critical Data and Algorithmic Sovereignty" provision of China's updated National Security Law. According to state media and regulatory filings, the agency concluded that integrating Manus's proprietary agentic AI models into Meta's global infrastructure would expose sensitive domestic user data and strategic algorithmic architectures to foreign intelligence. The regulatory body stated that Manus's deep integration into Chinese enterprise workflows made its foreign ownership an unacceptable vulnerability to the state.[4]

Manus is widely considered a pioneer in "agentic AI"—systems capable of autonomously executing complex, multi-step tasks across different software applications without human prompting. Prior to the acquisition's close in March, Manus had deployed its enterprise agents across thousands of Chinese firms, automating supply chain logistics, financial auditing, and software development. Meta had intended to use Manus's framework and engineering talent to accelerate its own autonomous agent rollout across WhatsApp, Facebook, and its enterprise offerings.[2][5]

Meta has stated it is "deeply disappointed" by the ruling and is currently evaluating its legal and regulatory options in both jurisdictions. The unwinding process presents a severe financial and technical headache for the Silicon Valley giant. Meta had already begun integrating Manus's engineering teams, proprietary datasets, and model weights into its Fundamental AI Research (FAIR) division. Reversing the deal will likely incur hundreds of millions in unwinding fees and severance, while forcing Meta to scrub Manus's code from its upcoming Llama 4 agent infrastructure.[1][2][4]

Western analysts broadly view the NDRC's intervention as direct geopolitical retaliation for recent US policies targeting Chinese technology. Over the past two years, Washington has forced the divestment of TikTok, banned the export of advanced AI accelerators to Chinese firms, and mandated "eyes off" data privacy for federal AI contracts. Blocking the Manus deal allows Beijing to demonstrate reciprocal leverage, proving it can inflict equivalent pain on American tech champions seeking access to Chinese innovation and engineering talent.[2][3]

The legal mechanism used by the NDRC relies on a novel interpretation of China's 2024 AI Governance Framework. While previous regulations required pre-approval for foreign acquisitions of critical physical infrastructure, the NDRC is now classifying foundational AI models and agentic frameworks as "strategic sovereign assets." Legal experts note that this reclassification effectively subjects all future AI-related mergers and acquisitions in China to national security vetoes, regardless of the target company's size, sector, or initial regulatory clearance.[4]

It remains highly uncertain how the unwinding will be enforced if Meta resists or claims technical impossibility. While the NDRC can easily block Manus's operations within China and seize its domestic servers, Meta already possesses the startup's core weights and source code on US soil. If Meta refuses to delete the transferred intellectual property, Beijing could retaliate by sanctioning Meta's hardware supply chains, particularly its reliance on Chinese components for its virtual reality headsets and data center networking gear.[1][3][5]

The collapse of the Meta-Manus deal is already forcing global venture capital to rapidly reassess the viability of cross-border AI investments. Several pending deals between European tech firms and Chinese AI labs were reportedly paused this weekend as compliance teams digested the NDRC's directive. The ruling cements a bifurcated global AI ecosystem, where American and Chinese AI development will proceed in strictly isolated silos, governed by competing national security mandates rather than open-source collaboration.[2][3][5]