Apple Intelligence Cleared in China With Alibaba Qwen

Apple Intelligence has cleared China’s AI rules

Apple Intelligence has been registered for public use in China, clearing a major regulatory hurdle that has kept Apple’s AI suite out of the country since it was announced in 2024.

China’s Cyberspace Administration added Apple’s service to its approved list on 15 July 2026. The registration covers Apple Intelligence features across iOS, iPadOS, macOS and visionOS, with Alibaba’s Qwen model set to handle text processing, image work and content generation for Chinese users.

The approval is essential rather than procedural. China requires generative AI services, including those running on devices, to be registered before they can be offered publicly. Foreign firms also need to work with domestic AI providers instead of deploying their own models independently, forcing Apple to find local partners for its China rollout.

Apple has not publicly commented on the regulatory listing. Alibaba, however, has confirmed its role. “Qwen will be integrated into Apple Intelligence experiences,” an Alibaba spokesperson told CNBC.

Alibaba is the headline partner, but Baidu is involved too

Alibaba’s statement makes Qwen the clearest named technology behind the Chinese version of Apple Intelligence. According to reporting by TechCrunch and The Next Web, Qwen will be incorporated inside Apple’s existing interface, rather than requiring users to open a separate Alibaba app. That means Chinese customers should be able to use text and image features through Apple’s own software experience.

But this is not an Alibaba-only arrangement. Reuters reported that Apple’s Chinese AI features will draw on models from both Alibaba and Baidu. Alibaba was previously reported to be building the primary system, while Baidu would play a smaller role. Baidu has also said it is working with Apple on Apple Intelligence features for Chinese iPhone users.

The setup reflects the particular constraints Apple faces in China. The company was reportedly exploring an agreement with Baidu before turning to Alibaba, after encountering challenges adapting models for Chinese customers. In mid-June, Alibaba unveiled its latest AI model and said it was compatible with Apple Intelligence, pointing to a deal that was nearing readiness.

Apple appears to have spent considerable time resolving the technical and compliance issues. Apple Intelligence was briefly enabled in China in March, apparently by mistake, but the formal registration now ends the longer process of gaining official approval.

When will Apple Intelligence arrive in China?

There is no confirmed launch date. The regulator did not give one, and Apple has not announced one either. Still, approval normally comes only a few months before a service becomes available, so a rollout around Apple’s usual autumn software releases is plausible.

Apple’s new platforms, including iOS 27, are currently in beta. A public release is expected around September, although Apple has not confirmed a precise date. If the Chinese version is ready in time, this approval gives Apple a route to launch the service alongside that broader software cycle.

For Apple, the timing matters. Greater China sales rose 28% to $20.5 billion in the second quarter, while IDC figures cited in reporting show iPhone shipments in China climbed 24.4% year on year in Q2 2026. That made Apple the fastest-growing major smartphone brand in a market where total shipments fell 4.3%.

Apple has also regained the number-two position in China’s smartphone market after a recent shopping festival brought discounts to the iPhone range. Giving customers locally approved AI features could help the company maintain that momentum against domestic rivals that already offer their own generative AI services.

Why Qwen’s on-device potential is worth watching

Qwen is Alibaba Cloud’s family of AI models, with support for 119 languages and dialects and variants designed for text, vision, audio and agent-style tasks. Alibaba says its Apple integration will enable on-device capabilities without routing data externally.

That promise is particularly notable because running sophisticated models locally remains difficult on a phone. PrismML, a Caltech spinout backed by Khosla Ventures, released compressed versions of Alibaba’s open-source Qwen model on 14 July. The company said it had reduced a roughly 54GB model to under 4GB, allowing all 27 billion parameters to run on an iPhone 15 or newer.

PrismML says the compression can reduce memory use by more than 90%, raise inference speed by six to eight times and lower power consumption by three to six times. Its chief executive, Babak Hassibi, told CNBC that Apple and other companies have been evaluating the startup’s models for speed, energy efficiency and on-device compatibility.

That does not mean Apple has struck a deal with PrismML. No formal partnership has been announced, and Apple may be pursuing its own technical approach. Still, the work illustrates why Apple’s China rollout could have implications beyond one market: bigger local AI models could eventually reduce reliance on cloud computing while strengthening privacy.

What it means for UK iPhone owners

There is no immediate change for UK buyers. Apple Intelligence is already available in Europe, while China is catching up after its separate regulatory process. However, Apple has said that some new Apple Intelligence and Siri AI capabilities are delayed in the EU for iOS 27 and iPadOS 27 because of the Digital Markets Act.

The UK is not subject to the EU’s Digital Markets Act, but Apple has not set out a distinct UK availability plan from its wider European approach. For now, the China approval is more important as evidence that Apple is willing to adapt Apple Intelligence to regional rules and local technology partners.

Apple Intelligence and Siri AI require an iPhone 15 Pro or newer with an A17 Pro chip and 8GB of RAM. The longer-term question for UK customers is whether Apple can bring more capable features onto the device itself, making AI faster, more private and less dependent on remote servers.