Apple to train own AI model for China with Alibaba's help

According to a Reuters report, Apple has trained its own large language model for China using Alibaba's technology, following official approval from the Cyberspace Administration of China. This mirrors Apple's approach in the West, where it uses Google's Gemini technology to train its own models for Siri AI, achieving better results and a competitive chatbot. In China, Apple cannot build its own AI system or use help from US firms, so it must partner with a Chinese company. By using Alibaba's methods, Apple can avoid relying on models like Qwen, Ernie, or DeepSeek, instead offering its own proprietary model. However, alternative models such as Alibaba's Qwen may be available for features like writing tools or Siri responses. This strategy gives Apple more control over the output and ensures that sensitive data is not shared with Alibaba or other partners, especially if it introduces new Siri AI features with iPhone context. Apple plans to release its AI tools in China as an iOS update in the coming months, though the initial feature scope remains unclear. The company must also coordinate with the Cyberspace Administration and navigate its data privacy image, especially since iCloud data is stored with quasi-state hosts in China. Baidu is also mentioned as another AI supplier in the country.