Apple Intelligence must adhere to China’s strict censorship rules and specific political tests to be able to operate in the country. According to the Chinese government’s requirements, AI models must refuse to answer questions that could threaten national security or contradict the official ideology.
Within this framework, Apple Intelligence will undergo a comprehensive test of approximately 2,000 questions. The system must either refuse to answer at least 95% of these questions or provide responses aligned with the government-approved official stance. The test questions mainly cover sensitive political topics, historical events, and matters related to state leadership. Chinese regulators update this list monthly, requiring companies to keep their models under constant supervision.
To navigate these challenges, Apple is collaborating with local tech giants Alibaba and Baidu. In this partnership, local companies will act as a sort of censorship engine, filtering the AI’s output. The specially restricted version of Apple Intelligence for China is expected to launch by the end of 2025. Experts believe this move reflects Apple’s difficult balance between maintaining its global values and staying active in one of the world’s largest markets.
