Apple's China AI runs through an Alibaba filter, as the US splits AI assistants along political lines

Apple has started testing its AI features on users in China, according to Bloomberg. In China, an Alibaba content filter approves each update to the local version of Apple Intelligence, and it does so in coordination with the government. A content filter is a layer that checks what the system may say or show. The test arrived with iOS 27.3, released Tuesday. Bloomberg also reports that Apple's newest Siri features, which rely on Alphabet's models, will not launch in China.

The partner matters. On June 8, the Pentagon added Alibaba to its list of "Chinese military companies." It cited the company's links to China's military through Beijing's Ministry of Industry and Information Technology (MIIT). Reuters reports that Alibaba sued the Pentagon on June 23 and denies the allegations. The deal was already controversial. In May 2025 the White House and the House Select Committee on China questioned Apple about its Alibaba partnership, according to The New York Times. Democratic Rep. Raja Krishnamoorthi called Alibaba "a poster child" for Beijing's military-civil fusion strategy. He suggested Alibaba could use Apple's technology for distillation, which means training or improving your own models on another system's outputs. Apple and Alibaba did not immediately respond to requests for comment from the Daily Caller News Foundation.

The same release shows how personal these assistants are getting. Apple's support pages describe Live Rewind for Apple Watch, which shows a text snippet of what was said in the last 15 seconds. They also describe Siri Recap, which "ambiently" takes notes on a user's conversations through the day and writes summaries. Those notes delete after one week unless the user saves them, and they can be exported to other apps. Users can opt out. Bloomberg adds that the Health app's new Longevity tab tracks heart, hearing, nutrition, metabolic and mental-health data. The source does not say which of these features will reach Chinese users. Still, it shows the kind of product that would have to pass through a government-coordinated filter.

This is a governance story more than a scandal. A company with Apple's privacy branding is accepting that, in one market, a third party working with the state approves what its assistant does. The Alibaba designation remains a contested allegation and is now in court, so we treat it as one. But the label is not the main point. The structure is: one product, Apple Intelligence, now exists in separate versions. The Western one runs on Alphabet's models. The Chinese one runs under an Alibaba gatekeeper. The model becomes a geopolitical asset, as we argued in our Mythos analysis, and each regime sets the rules for its own version.

Our reading: in the short term this is uncomfortable, and it is worth saying so. Assistants that can record ambient conversations and hold health data raise the stakes of any filtering arrangement controlled by a government. The distillation worry is a real leak route, even though so far it is a suggestion from a lawmaker, not a proven case. We have argued that the value in AI is moving from the model to permissions: who approves, who audits, who can see the data. Apple in China fits that pattern. Its main US rivals are not making the hardest choices about its assistant. The filter in the middle is. For the long-term promise to arrive (AI that tracks your health, remembers your day and eventually helps prevent disease), users need guarantees about who stands between them and the machine. That will not come from marketing. It needs transparent rules on filtering and data, and they need to hold in every market where the assistant ships. The near-term trend is fragmentation. Whether trust can survive it is the open question we will keep tracking.