How iOS 10 will collect and use your data

Apple believes there shouldn’t be a tradeoff between intelligent software and privacy. iOS 10 is the first to incorporate machine learning and artificial intelligence without sacrificing privacy.

[background music]
Man: IOS 10 marks Apple's biggest push into machine learning ever, but that doesn't mean your iPhone will have to expose your deepest, darkest secrets.
[background music only]
Man: Unlike Google and Facebook, Apple never had to collect your data in order to make a profit. Today, big data is the driving force behind machine learning and artificial intelligence, technology that Apple realizes can make software more helpful, more proactive, and more like Scarlett Johansson.
[movie dialog]
Man: Machine learning actually has a lot of useful applications. Just think about it. We all started messaging words like bae, turnt, and FOMO way before they ended up in the dictionary.
This technology could've saved us from some very awkward autocorrects. There's a big difference between texting that you're in a FOMO coma and texting that you're in a FEMA camp.
For machine learning to really take off, it needs to aggregate as much user data as possible in order to figure out useful patterns. That means your data, too. Thankfully, Apple seems to have figured out the magic formula with iOS 10, so there won't be a need for a trade off between privacy and smarter devices that rely on this machine learning.
With iOS 10, Apple introduces the concept of differential privacy, a way of collecting big data while implementing statistical distortions so that your individual information remains private.
According to Apple, iOS 10 uses this to improve the QuickType keyboard, Spotlight search, the Notes app, and emoji suggestions, because now we need to leverage all the world's knowledge to help us figure out just the right emoji.
Apple doesn't care to know whether I got turnt last weekend, but it wants to know whenever a new word becomes hot lingo. Apple's differential privacy is about spotting those big picture patterns, not the nitty gritty details to sell me a hangover cure.
Beyond differential privacy, Apple's relying on the iPhone itself to run a lot of the machine learning computations, which means our data doesn't venture outside the iPhone's encrypted walls to some server in the middle of nowhere.
For example, iOS 10 gives the iPhone the power to analyze our entire Camera Roll and pick out our friends to create albums based on facial recognition. This all happens locally on the device, however, so your selfies are never cross referenced against photos on other iPhones.
Even though iOS 10 is the gateway to the wide world of machine learning, no one has to know just how much you love KIMOJI.