Apple unveiled a wave of AI features at WWDC24, promising powerful tools for iPhones, iPads, and Macs. But can Apple truly guarantee privacy in an AI world? Their on-device processing and private cloud solutions are impressive, but the integration of ChatGPT creates a potential loophole.

How Apple Plans to Make Its AI Private

Apple revealed many AI features at WWDC24, and they will be available on all supported Macs, iPhones, and iPads later in 2024. However, AI and privacy don’t often go hand-in-hand. Regardless, Apple has made bold but reasonable commitments to guarantee privacy.

On-Device Processing

Typically, AI operations demand substantial processing power, a resource that most standard phones and laptops lack. However, with the advent of Apple silicon, all Macs, iPhones, and iPads equipped with the M family of chips or A17 Pro (and later) will be capable of executing most of Apple’s generative AI models locally, including Apple’s language model.

Apple has been developing its proprietary Apple Neural Engine (ANE) to handle machine learning and AI processing for a while now. Over the years, it has become more efficient in processing AI requests without leaving them entirely to the CPU or GPU. Most of Apple’s AI models will truly run offline and ensure privacy.

Private Cloud Compute (PCC)

However, your Apple device may be unable to handle some AI requests locally. While Apple stated this in WWDC24, it didn’t specify or give examples of what requests would be too much for the local offline processors. Examples are up to speculation, but Apple says “complex requests” that would otherwise bog down your device.

Apple’s solution is Private Cloud Compute (PCC), a cloud-based AI processor built with custom Apple silicon. It places a strong emphasis on privacy and transparency and comes with its own Secure Enclave, just like the one that protects your Apple device. When you use PCC, your AI requests are immediately deleted after processing, and no one, including Apple, can access the data through a backdoor. To further build trust, Apple offers verifiable transparency, allowing third-party security researchers to examine the software running on PCC nodes.

ChatGPT Poses the Primary Privacy Risk

On-device processing and PCC are both great privacy stanchions, but there’s still one aspect of Apple Intelligence that is questionable: ChatGPT. For those who don’t know, Apple plans to integrate ChatGPT into its AI experience, including Siri. These AI applications will tap into ChatGPT’s broader knowledge base and borrow its image-processing capabilities.

To be completely fair, Apple has taken some measures to guarantee your privacy while using ChatGPT. For instance, when you make a request to OpenAI, your device will obfuscate your IP address, and a profile storing information about you won’t be created. Additionally, all ChatGPT requests require express permission from you beforehand, and you must grant permission each time.

Despite Elon Musk’s claims to the contrary, OpenAI can’t access all the data on your device; ChatGPT can only process the information presented in its prompts.

However, ChatGPT's most powerful features are behind its subscription paywall, and you’ll have to connect to your premium account to enjoy them. Doing this, however, will make you subject to OpenAI’s privacy policy, which isn’t as generous as Apple’s, to put it mildly. OpenAI’s privacy policy allows it to collect personal information, log usage, train models on personal information, share your data with third parties, and many other concerning implications. You can view it on OpenAI’s website.

While there are ways to opt out of some of ChatGPT’s privacy breaches, most solutions are roundabout, and OpenAI can easily circumvent them. It also seems that avoiding ChatGPT on Apple Intelligence (especially on macOS) will be difficult, if not impossible. Depending on the way you use Siri or Compose, you might find yourself heavily reliant on ChatGPT, leaving much of your data in the hands of OpenAI.