While Meta’s consumer platforms like Facebook and Instagram have been blocked by China’s “Great Firewall” since 2009, the tech giant has spent the last few years quietly—and inadvertently—poking massive holes in that same digital barrier. Not with social media, but with artificial intelligence.
For years, Meta championed an open-source (or open-weight) strategy with its highly successful “Llama” series of AI models. The goal was to foster global innovation, developer mindshare, and democratize access to advanced AI. However, this open approach provided Chinese technology companies and researchers—such as Alibaba and Tencent—with unrestricted access to state-of-the-art AI models that they could adapt, fine-tune, and deploy on their own domestic platforms.
Essentially, high-level foreign AI technology was able to bypass the traditional restrictions that usually limit the flow of foreign digital services into the Chinese market.
The Rise of Chinese Open-Source AI
The impact of this access has been profound. While Silicon Valley companies initially built the open-source playbook, Chinese labs leveraged this openness to rapidly accelerate their own AI capabilities.
By early 2026, the landscape shifted dramatically. Chinese models, particularly Alibaba’s “Qwen” family, became dominant in terms of global open-source downloads. In many benchmarks, these domestic models began outperforming their Llama counterparts, demonstrating how effectively Chinese companies had learned from and built upon the foundations laid by Meta.
This dynamic has created a complex geopolitical irony and raised significant concerns in the U.S. about the security and competitive implications of freely sharing frontier technology with geopolitical rivals.
A Strategic Pivot: The Launch of Muse Spark
Perhaps recognizing the shifting balance of power and increasing regulatory scrutiny, Meta is now changing its playbook. In April 2026, Meta signaled a departure from its purely open-source path by launching Muse Spark, its first major closed-source AI model.
This move toward a more proprietary, hybrid strategy suggests that the era of freely available, frontier-level AI models from Western tech giants might be drawing to a close as the global AI race intensifies.
The Geopolitical Pushback
The tensions aren’t just limited to open-source code. Just as Meta attempts to tighten its grip on its intellectual property, Beijing is doing the same.
Also in April 2026, the Chinese National Development and Reform Commission formally blocked Meta’s $2 billion acquisition of Manus, a Singapore-based AI startup with Chinese roots. Beijing viewed the acquisition as a potential “technology leakage” of homegrown Chinese innovation and talent to a major U.S. company. This rare move to unwind an already-signed corporate deal underscores the incredibly complex and restrictive regulatory environment surrounding data, AI, and foreign investment in the region.
Ultimately, Meta’s initial open-source strategy may have successfully accelerated global AI development, but it also empowered formidable competitors behind the Great Firewall. As Meta pivots towards closed-source development with Muse Spark, the next phase of the global AI race promises to be far more guarded.