Is Apple Intelligence safe to use?
Apple’s built-in personal AI system on supported iPhone, iPad, and Mac. It powers features like Writing Tools, notification and message summaries, image features, and deeper Siri help. Many tasks run on the device; harder ones can use Apple’s Private Cloud Compute servers. An optional ChatGPT extension can hand some requests to OpenAI — that path is separate and off by default. Apple Intelligence is not available on every device, language, or region.
Last reviewed: 24 August 2026
Does Apple Intelligence train on what you type?
No.
Can you turn Apple Intelligence's training off?
See what left the device: Settings → Privacy & Security → Apple Intelligence Report → choose the last 15 minutes or 7 days → Export Activity. Open Apple_Intelligence_Report.json if you want to see Private Cloud Compute and ChatGPT extension requests. The report can be empty if there have not been any recent off-device requests. Device Analytics: Settings → Privacy & Security → Analytics & Improvements → turn off Share iPhone Analytics or Share iPad Analytics. On Mac, use System Settings → Privacy & Security → Analytics & Improvements → Share Mac Analytics. ChatGPT extension: Settings → Apple Intelligence & Siri → ChatGPT → turn off Use ChatGPT. If you keep it on, leave Confirm ChatGPT Requests on so Siri cannot send quietly. If you signed into ChatGPT through the extension, also check ChatGPT’s own Data Controls, including model-improvement and temporary-chat settings. Those rules are OpenAI’s, not Apple’s. Whole feature: you can turn Apple Intelligence off in Settings if you want no on-device or Private Cloud Compute help. That is the blunt instrument. Most people only need the ChatGPT switch, the analytics switch, and the report.
Is what you put into Apple Intelligence private?
On your device first. Apple Intelligence is built to use personal context — calendar details, apps you use often, what is on the screen — without Apple building a cloud profile of that private life. When the on-device model can finish the job, such as many email, message, and notification preview summaries, the work stays on the phone or computer. Private Cloud Compute is still off-device, but it is not a normal chat-log cloud. If a task needs a bigger model, Apple’s own example is proofreading or rewriting an email with Writing Tools, the device may send only the data needed for that request to Apple silicon servers in Private Cloud Compute. Apple’s privacy pages say that content is processed to answer you, is not stored by PCC, and is not made accessible to Apple for reading later. Your device is designed to talk only to servers whose software has been publicly logged so outside experts can inspect it. Apple may keep limited operational stats, such as rough request size, which feature was used, and how long the request took. Those stats are not the text of your email and are not tied to your Apple Account. Training is the cleaner part. Apple says it does not use users’ private personal data or their Apple Intelligence interactions to train its foundation models. If you opted into Share Device Analytics, Apple may use privacy-preserving aggregated trends, including about Apple Intelligence, to improve features. That is not the same as training foundation models on your private chats, but if you do not want even that, turn it off under Privacy & Security → Analytics & Improvements. The ChatGPT extension is the important third door. It is off by default. If you turn it on, some requests and attachments can go to OpenAI, not to Private Cloud Compute. Without a ChatGPT account, Apple and OpenAI say OpenAI processes the request to answer you, does not store it unless legally required, and does not use it to train models. If you sign into a ChatGPT account, OpenAI’s account settings and privacy policies apply; it may log requests, attachments, and history, and may use that data to improve or train models depending on your settings. That is where the privacy story changes. Very small print, annoyingly large consequence.
So, is Apple Intelligence safe to use?
Generally, yes. We currently rate Apple Intelligence "Generally Safe" (privacy risk: Low), based on how it handles the data covered above.
The Skeptic's take
“I kept saying “it all stays on my iPhone,” like that was one simple sentence. It isn’t. A lot does stay on the phone. The hard stuff can go to Apple’s private servers that aren’t supposed to keep a copy. And if I tap ChatGPT, I’ve left the building. I exported the Apple Intelligence Report once so I’d stop guessing.”
— The Skeptic
The official documents behind this review
We read these so you don’t have to. Check them yourself any time.
Wondering how we decide these ratings? Read how we review AI privacy.
Knowing what’s safe is half of it.
The other half is knowing what to do with these tools. We teach that part — tutorials, prompts, plain English, no judgment. For less than a cup of coffee a month.
Show Me How This Works →Start free. Cancel anytime. No judgment, ever.