China's National Development and Reform Commission issued a formal order on April 22 requiring Meta to divest its controlling stake in Manus, the Singapore-based AI startup it acquired for $2.3 billion in January. The NDRC determination cited concerns over data flows, algorithmic accountability, and "unacceptable concentration of AI capabilities in foreign-controlled entities with broad consumer access."

The acquisition had moved quietly through regulatory review in Singapore and the EU but faced sustained scrutiny from Beijing from the moment it was announced. What changed in the past 90 days, according to sources familiar with the NDRC's internal deliberations, was Meta's integration of Manus's agentic task-execution technology into its consumer AI products "',"Å",¾ specifically a set of autonomous agents embedded in Meta AI that can browse, book, and transact on behalf of users. That integration appears to have crossed a threshold that China's regulators had flagged but not yet acted on.

What the Order Actually Targets

The NDRC's language matters. It doesn't frame this as a national security issue in the traditional sense "',"Å",¾ no talk of military applications or surveillance infrastructure. Instead, it focuses on "algorithmic accountability" and "consumer AI concentration." That framing is deliberate: it positions China as a regulator protecting its domestic AI ecosystem rather than as a purely geopolitical actor. The effect, however, is the same "',"Å",¾ Meta must unwind a major cross-border AI acquisition under regulatory duress.

Manus's technology is now effectively orphaned in Meta's portfolio. The company's agentic AI stack "',"Å",¾ which Meta had planned to integrate across WhatsApp, Instagram, and its AR glasses "',"Å",¾ faces a forced divestiture timeline that sources put at 180 days. Several major institutional investors in Manus, including Sequoia Asia and Temasek, are now parties to a dispute over the valuation and unwind mechanics of the original deal.

Implications for Global AI M&A

The Manus unwinding is a data point in a larger pattern. Over the past 18 months, cross-border AI acquisitions have faced an unprecedented level of regulatory friction on multiple fronts simultaneously "',Å",¾ US export controls, Chinese regulatory review, EU merger rules, and now a new layer of algorithmic accountability scrutiny. The message to the market is that AI acquisitions involving entities with data access to large consumer bases will face heightened review regardless of where the acquiring company is headquartered.

For dealmakers, this is a structural change, not a one-off. Founders building AI companies with global ambitions now need to think about acquirer jurisdiction from day one, not just at signing. The Manus story will be studied in tech policy classrooms for years "',"Å",¾ not as a cautionary tale about Meta's deal process, but as the moment cross-border AI M&A stopped being routine.