Apple Intelligence is finally coming to China — but it won't be running Apple's own model.

On July 15, China's cyberspace administration announced that 7 mobile end-side generative AI services, including Apple Intelligence, have completed regulatory filing. The most closely watched one, Apple Intelligence, will run on Alibaba's Qwen large language model for the Chinese market. Also filing on the same day: Huawei Xiaoyi AI, OPPO Andes GPT, vivo Blue Heart End-Side LLM, Xiaomi HyperMind AI, Samsung Galaxy AI, and Nubia Doubao Phone LLM.

Seven products, all approved on the same day. This isn't a coincidence — it's end-side AI regulation officially going live. From now on, AI features in phones can't just be shipped whenever you want — you have to pass the filing process first.

Why Is Apple Using Alibaba's Model?

Many people's first reaction: Apple is so powerful, why not use its own model?

The answer is simple: regulatory requirements. China's generative AI regulations state that LLMs providing services to the Chinese public must be trained in China, deployed in China, and data must stay in China. Apple Intelligence's core model is developed with Google Gemini, with training data and deployment all overseas — bringing it directly to China wouldn't meet regulatory requirements.

Why doesn't Apple just train its own model in China? Because the cost and timeline don't make sense. Training a high-quality end-side LLM requires massive amounts of Chinese data, compute, and time. Rather than building from scratch, it's smarter to find an already-compliant, already-mature local partner and integrate directly.

This is obviously great news for Alibaba — becoming Apple's China AI supplier is like getting a "premium customer endorsement." From now on, when Qwen goes after other clients, it can just say "Apple uses us."

But the deeper meaning is this: globalization in the AI era is completely different from the internet era. In the internet era, you launched software globally and at most did some local translation. Not so with AI — every country has its own data sovereignty requirements, content regulations, and security review requirements. To enter a market, you might need a completely different model.

What Does End-Side Filing Mean?

Seven mobile end-side AIs filing on the same day is an even more important signal than Apple using Qwen.

Until now, most people's understanding of AI regulation was focused on "cloud LLMs" — services like ChatGPT, Wenxin Yiyan, and Doubao all need to file. But end-side AI is different — the model runs locally on your phone, data doesn't get uploaded to servers — so why does it need filing?

Because the regulatory logic has shifted. It used to be about "where data is stored"; now it's about "what content is being output." Whether the model is in the cloud or on-device, as long as it generates content and interacts with users, it falls under regulatory scope.

What does this mean? It means every AI feature on your phone — AI photo albums, AI writing, AI photo editing, AI voice assistants — will need regulatory review and filing before it can launch. Phone makers can no longer just stuff a bunch of AI features into a system update like they used to. Every feature has to be reviewed first.

For consumers, this might mean slower feature rollouts and fewer "magic features." But for the industry as a whole, it's the necessary path from wild growth to standardization.

Apple using Qwen to enter China is just the beginning. In the future, more and more global tech companies will partner with different local AI vendors to enter different national markets. Globalization isn't dead — but its form has changed: from "one product for the whole world" to "one framework, many local models."

See you tomorrow.