Apple is quietly changing how it will deliver artificial intelligence in its biggest and most tightly regulated overseas market. According to multiple reports published August 14, the company has trained a custom large language model for China with support from Alibaba — a departure from Apple's previous approach of relying primarily on models developed by Chinese partners such as Baidu and Alibaba's Qwen.
The shift matters because services like OpenAI's ChatGPT are unavailable in mainland China, and China requires generative AI services offered to the public to pass regulatory review. Apple has already registered its on-device generative AI service with Chinese regulators, clearing an important hurdle ahead of a planned Apple Intelligence rollout in the country. Alibaba has confirmed that its Qwen model will be integrated into Apple Intelligence across iOS, iPadOS, macOS and visionOS for Chinese users.
Training its own model would give Apple greater control over the AI experience on devices sold in China — and a proprietary alternative to lean on as regulatory and geopolitical pressures fragment global technology stacks. It also sharpens Apple's competitive position against Huawei and other domestic smartphone makers that have made AI central to their pitch.
Analysts see the move as a case study in AI fragmentation: global companies increasingly need different models, infrastructure partners and compliance strategies for different markets. For Apple, the China model is both a regulatory necessity and a hedge — one that lets the company keep its AI destiny in its own hands even where the frontier models it relies on elsewhere cannot reach.
