Meta’s $2 billion Manus AI deal is reportedly being pulled apart after Beijing demanded the transaction be unwound.

The company has completed an operational split from Manus and stopped data sharing between the two firms, according to TechCrunch, citing Bloomberg. The move follows China’s earlier order for Meta to undo the acquisition on national security grounds.

Meta’s Manus deal is turning into an AI power struggle

Meta bought Manus to strengthen its push into AI agents — tools that don’t just answer questions, but can complete tasks, automate workflows, and act across apps.

That makes Manus more than another chatbot company. Its own website describes the product as an “action engine” built to execute tasks and automate workflows.

That kind of technology matters to Meta because it wants AI inside Facebook, Instagram, WhatsApp, Messenger, and business tools. We’ve already seen Meta push deeper into AI-powered commerce, including its plans for WhatsApp AI agents that could help businesses handle customer chats and sales.

But this deal now shows the harder part of the AI race.

The fight isn’t only about who has the best model. It’s about who controls the talent, data, code, and infrastructure behind those models.

Why Beijing stepped in

Chinese regulators reportedly ordered Meta to unwind the Manus deal in April 2026, citing national security concerns. Reuters reported that China’s National Development and Reform Commission became involved because the deal touched sensitive technology and Chinese-origin AI talent.

That matters because Manus had Chinese roots, even though it later operated from Singapore.

Beijing appears to be sending a clear message: moving a company’s headquarters abroad doesn’t automatically move its technology outside China’s reach.

Here’s the simple version:

Issue	Why it matters
AI agents	They can automate real work, not just chat
Chinese-origin founders	Beijing may still claim oversight
Foreign acquisition	US ownership raises national security concerns
Data sharing	Regulators worry about technology transfer
Meta integration	Once tools merge, separation gets messy

Reuters also reported earlier this year that China would assess Meta’s acquisition of Manus and warned companies to comply with Chinese laws around foreign investment, tech exports, data transfers, and acquisitions.

So this didn’t come from nowhere.

Meta now has to separate what it wanted to combine

The most difficult part is not announcing a split. It’s proving that the split actually happened.

According to reports, Meta has now stopped data sharing with Manus and separated operations. TechCrunch described the move as the clearest step yet toward complying with Beijing’s divestiture demand.

That sounds simple. It isn’t.

When a large tech company buys an AI startup, it usually wants three things:

• The people who built the product
• The models and systems behind it
• The product roadmap that can plug into bigger platforms

If engineers, code, infrastructure, and product ideas have already started moving across teams, regulators may ask a tough question: what exactly can be unwound?

That’s the expensive part.

Meta may not only lose access to Manus technology. It may also lose time in a race where OpenAI, Google, Anthropic, Apple, and Microsoft are all pushing AI deeper into everyday tools.

Why this matters beyond Meta and China

This story may sound like a Silicon Valley problem. It’s bigger than that.

AI has become strategic infrastructure. Governments now treat powerful models, chips, data, and engineering talent like national assets. That means regulators can block or reverse deals even after companies think the paperwork is done.

For investors, this raises the risk around AI startups with links to China, the US, or other politically sensitive markets.

For founders, it creates a new question: can your startup really sell to anyone, anywhere?

And for users, including those in South Africa, it reminds us that the AI tools we use daily often depend on global deals we never see.

A small business in Johannesburg using WhatsApp automation, a creator in Cape Town relying on Instagram tools, or a retailer testing AI customer support may feel the impact later if Meta’s AI roadmap slows or changes.

The South African angle: AI dependence comes with hidden risk

South Africa doesn’t sit at the centre of the US-China AI fight. But we still feel the aftershocks.

Meta owns platforms that many South Africans use for communication, marketing, sales, and customer support. WhatsApp, Facebook, and Instagram are not just social apps here. They are business infrastructure.

That means Meta’s AI strategy can shape how local businesses handle customer service, ads, payments, and product discovery.

If major AI features depend on cross-border acquisitions, then geopolitical tension becomes a product risk. Your chatbot update, ad tool, or automated support feature may depend on whether regulators in Beijing, Washington, or Brussels allow a deal to stand.

That’s the lesson.

AI may feel borderless when you open an app. Behind the scenes, it’s becoming more controlled, more political, and more expensive to scale.

Meta wanted Manus because AI agents could make its platforms more useful and more profitable. Beijing pushed back because that same technology may be too strategic to let go.

FAQs

What is Manus AI?

Manus is an AI agent startup that builds tools designed to complete tasks, automate workflows, and act beyond simple chatbot replies.

Why did China reportedly demand the Meta-Manus deal be unwound?

Chinese regulators reportedly cited national security and technology transfer concerns, especially because Manus has Chinese roots and works in advanced AI.

Why should South Africans care?

Meta’s AI decisions affect tools many South Africans use every day, especially WhatsApp, Instagram, and Facebook. If global AI deals stall, local product features can change too.