Does Apple Dictation Go to the Cloud?

Yes — sometimes. Apple's own documentation states that Dictation requests “can process data on-device or on Apple servers.” On recent hardware with supported languages, basic dictation stays local. But in many scenarios — older devices, certain languages, search fields, Siri integration — your voice may travel to Apple's servers.

The issue isn't that Apple is hiding something. They're transparent about this in their privacy documentation. The issue is that there's no user-facing indicator telling you which path your audio is taking right now.

When Apple Dictation stays on your device

Apple has invested heavily in on-device machine learning, and it shows. On supported hardware, Dictation runs locally — no internet required. The conditions for on-device processing:

• Apple Silicon Mac (M1 or later) or recent iPhone/iPad with Neural Engine
• Supported on-device language — Apple's list of on-device languages is shorter than its full dictation language list
• Standard dictation — not in a search field or Siri-triggered context
• Dictation enabled in System Settings → Keyboard

This is genuinely good engineering. Apple's on-device speech models are fast, accurate, and private when they're being used. The question is when they're being used.

When Apple Dictation goes to the cloud

Apple's on-device processing has clear boundaries. Your voice may be sent to Apple's servers when:

• Older hardware. Intel Macs and older iOS devices don't have the Neural Engine needed for on-device processing.
• Unsupported languages. Apple's list of on-device languages is much shorter than its full dictation language list. Many languages still require a server round-trip.
• Search fields. Apple's support page notes that when dictating into a search field, “text may be sent to the search provider.” This means your dictated text could reach Google, Bing, or another third party — not just Apple.
• Siri-integrated flows. Dictation that triggers Siri functionality may involve server processing even on supported hardware.

And crucially: there's no visual indicator. No icon changes. No notification. You have no way to confirm whether a given dictation session was processed locally or on a server.

What Apple does with server-side data

When your dictation does reach Apple's servers, here's what happens according to Apple's privacy documentation:

• Audio is associated with a random identifier — not your Apple ID.
• Data may be retained for up to six months to improve Siri and Dictation.
• After six months, the data may be kept without the random identifier for up to two years.
• You can opt out of contributing to improvements in Settings → Privacy & Security → Analytics & Improvements — but this doesn't prevent cloud processing, it only limits how Apple uses data it already received.

Apple's approach is better than many cloud services. But “better than most” and “private” are not the same thing.

The core problem: ambiguity

Apple Dictation's privacy story is conditional. It depends on your hardware, your language, what field you're typing in, and whether Siri is involved. For casual use, this is probably fine.

But if you dictate medical notes, legal communications, proprietary code, confidential business strategy, or anything you'd be uncomfortable seeing on a server you don't control — “probably fine” isn't the standard you need.

The absence of a guarantee is the problem. Not malice, not negligence — just ambiguity where certainty is required.

If you need a guarantee

Resonant processes speech recognition entirely on your Mac using local AI models. There is no cloud component, no server fallback, and no scenario where your audio leaves your device.

This isn't conditional on your hardware generation, your language, or what text field you're typing into. It's architectural — the same way ProtonMail can't read your emails because they never have the unencrypted data, Resonant can't access your voice because it never leaves your machine.

• Always local. No internet required. Works offline, on airplanes, in air-gapped environments.
• No data retention. Audio is processed in real-time and immediately discarded. We don't store recordings or transcripts.
• No ambiguity. There's no “on-device or server” conditional. It's always on-device.

Read the full technical breakdown of how Resonant handles your voice data →

Download Resonant for Mac →

Frequently asked questions