Apple’s use of differential privacy in iOS 10 to drive AI will be opt-in, limited to four use cases at launch

Apple is making many of its services smarter in iOS 10 and macOS Sierra through the use of differential privacy which enables artificial intelligence to become smarter while still balancing privacy. Given Apple’s strong position on protecting user privacy, Recode requested more details on how the company’s use of differential privacy would work. Apple so far has resisted collecting a lot of user data that could make Siri, iCloud, and other services smarter.

Its use of differential privacy to collect some user data to improve its AI services will be totally opt-in, according to the new report, and hasn’t been used in services before iOS 10 and macOS Sierra. Apple also reiterated that it isn’t collecting user data in iCloud-stored photos to improve the AI behind its subject recognition in Photos. That’s why indexing photos happens separately on each device (like Mac, iPhone, and iPad) and object/facial recognition data isn’t synced. Craig Federighi used the example earlier this month that Apple can recognize what a mountain is on their own without using customer photos.

Recode also shared that Apple’s use of differential privacy is mostly limited in scope at launch:

As for what data is being collected, Apple says that differential privacy will initially be limited to four specific use cases: New words that users add to their local dictionaries, emojis typed by the user (so that Apple can suggest emoji replacements), deep links used inside apps (provided they are marked for public indexing) and lookup hints within notes.

Here’s an example of Apple detailing how differential privacy works with deep links in its iOS 10 documentation:

iOS 10 introduces a differentially private way to help improve the ranking of your app’s content in search results. iOS submits a subset of differentially private hashes to Apple servers as users use your app and as NSUserActivity objects that include a deep link URL and have their eligibleForPublicIndexing property set to YES are submitted to iOS. The differential privacy of the hashes allows Apple to count the frequency with which popular deep links are visited without ever associating a user with a link.

Apple also detailed these efforts in its official press release for iOS 10:

Starting with iOS 10, Apple is using technology called Differential Privacy to help discover the usage patterns of a large number of users without compromising individual privacy. In iOS 10, this technology will help improve QuickType and emoji suggestions, Spotlight® deep link suggestions and Lookup Hints in Notes.

The report also clarifies other AI-driven efforts Apple already has underway that don’t rely on differential privacy at all. Proactive, for example, is a concept that matches much of what Google does through data collection but Apple relies on indexing data locally on the device without information ever being shared across the Internet. For example, iOS 9 started recommending calendar appointments based on information from email saved locally on your iPhone or iPad.