APREX · EST. 2026
After a lengthy wait and demanding negotiations with Chinese regulators, Apple Intelligence is finally ready for the Chinese market. Approval from China's Cyberspace Administration marks one of the most contentious expansions in Apple's recent history — and raises fundamental questions about whether the company can balance its global values against the demands of an authoritarian market.
Alibaba's Qwen takes over the engine role
Rather than running its own models or OpenAI's ChatGPT, as Apple does in the West, Apple Intelligence in China will be built on Alibaba's Qwen family of large language models. According to TechCrunch, the Qwen models will handle text and image understanding as well as generation across Apple's operating systems. Baidu is also expected to contribute to specific features such as Visual Intelligence.
Qwen is a rapidly growing model family with variants ranging from 0.6 billion parameters — designed for edge devices such as smartphones — to models with more than 480 billion parameters for more demanding tasks. The models have been recognized as serious competitors to offerings from Meta, OpenAI, and Anthropic across a range of benchmarks.
Censorship is not an add-on — it is the core
Approval did not come without a price. Chinese regulations for generative AI require Apple Intelligence to refuse to answer at least 95 percent of approximately 2,000 carefully crafted questions on politically sensitive topics. The list of prohibited subjects is broad: state authority, discrimination, pollution scandals, financial fraud, labor disputes, and political satire. Authorities update these question sets monthly, requiring ongoing adaptation of the models.
The local partners — Alibaba and Baidu — effectively function as a censorship layer between Apple's platform and its users, according to the research material reviewed by 24AI.
The local companies effectively act as a censorship engine that filters the AI's output on behalf of the authorities
A near-blunder in March 2026
The road to approval has not been without drama. In March 2026, Apple Intelligence was briefly rolled out in China by mistake — before official approval had been granted. The incident highlighted how seriously the authorities take the regulatory framework: companies risk administrative sanctions for breaching rules on AI security assessments, algorithm filings, and data protection.
Data protection under pressure
Apple markets its Private Cloud Compute technology globally as a guarantee that user data is processed without being stored. In China, the situation is different. Historically, Apple has stored Chinese iCloud data on servers operated by a state-owned company, with encryption keys held within the country's borders.
China's Cybersecurity Law and Cryptography Law grant the state broad access to encryption and decryption technology — something security experts and former Apple engineers have described as a de facto backdoor for government access to user data. These concerns apply not only to AI features but to Apple's entire data infrastructure in the country.
U.S. Congress investigates
The partnership with Alibaba has not gone unnoticed in Washington. A congressional investigation was launched in 2026 to examine the security risks of using Chinese AI models in American products — with Alibaba's Qwen explicitly named. The concerns center on both potential data access and the possibility that the collaboration strengthens the global position of Chinese AI providers.
For Apple, the China strategy represents what researchers describe as a strategic paradox: the country accounts for nearly one-fifth of the company's revenue, yet operating there requires concessions that potentially undermine the very values on which Apple builds its brand in the rest of the world.