Apple CEO pushes AI system launch in China after iPhone Air debut

Apple is working to bring its AI system, Apple Intelligence, to China, CEO Tim Cook said during an event in Shanghai on October 18.

Apple Intelligence, a built-in AI feature for iOS 18, iPadOS 18, and macOS Sequoia, has been unavailable on Apple devices in mainland China and Hong Kong due to local regulations.

Cook did not give a timeline for the China launch.

According to sources cited by the South China Morning Post, Apple has struck a deal to use Alibaba Cloud’s Qwen models for Apple Intelligence in China, while using OpenAI’s GPT models internationally.

During his six-day visit, Cook met with local officials, business partners, and developers, and committed Apple to increased investment in China during a meeting with officials.

Apple’s iPhone 17 series, launched last month, sold 14% more units than the iPhone 16 in the first 10 days in China and the US, according to Counterpoint Research.

🔗 Source: South China Morning Post

🧠 Food for thought

Implications, context, and why it matters.

Apple Intelligence in China depends on clearing the CAC’s generative AI filings

In China, the Cyberspace Administration of China (CAC) requires any generative AI that can influence public opinion or mobilize users to file with local cyberspace offices 1, which Apple Intelligence must follow.
By March 2025, 346 services had filed, including DeepSeek and Baidu’s Ernie Bot 1.
Launched apps must show model names and filing numbers in the interface 1, so an Alibaba Cloud Qwen integration would need that label in the UI or details pages.
Providers must keep training data free of IP violations and non-consensual personal data, keep audit logs, and risk fines up to RMB 100,000 or suspension for breaches 2.
South China Morning Post reported a February deal for Apple to use Qwen in China, with no timeline; approvals span security reviews, algorithm filings, and broader cybersecurity and data protection rules 2.

Enterprise software teams need dual model routing for China and the rest of the world

If Apple runs Qwen in China and GPT models elsewhere, B2B builders will need request routing by region, using account settings or IP geolocation.
Qwen’s API accepts OpenAI-style calls in the Singapore and Beijing regions 3, which lets teams swap backends without rewriting core logic.
Alibaba says Qwen supports 119 languages, with 600 million downloads and over 170,000 derivative models, which signals broad uptake 4.
Personal Information Protection Law (PIPL) rules apply 2. Keep minimal retention of identifiable data, avoid profiling or third-party sharing based on inputs or usage, plus a clear complaint channel.
Enterprise teams should use Qwen features like streaming output and function calling 3. Multimodal input and explicit caching help match features across regions 3.