Every time you ask an AI chatbot a question on your iPhone, your words leave your phone, travel to a server you've never seen, and get stored in a database you don't control.
Your health questions. Your financial worries. Your legal situations. Your relationship problems. Work documents with proprietary details. Drafts of difficult messages. Journal entries. Things you wouldn't post publicly.
All of it is sitting on someone else's infrastructure. And you have no real control over what happens to it next.
You assumed your iPhone kept things private. For your photos, your messages, your browsing, maybe it does. But the moment you use a cloud AI app, none of that protection applies. Your words go to a third party's server and Apple can't help you.
Your iPhone's Privacy Has a Gap
Your phone maker built its brand on privacy. But then they integrated a cloud AI chatbot directly into the operating system. When your phone's assistant can't handle a request, it routes your query to someone else's servers.
Their own on-device AI handles simple tasks locally. But for anything complex, it falls back to cloud servers. Sometimes their own. Sometimes a third party's. You don't control which path your query takes.
This isn't a flaw. It's just how cloud AI works. Your data has to go somewhere to get processed. And "somewhere" means a server owned by a company whose privacy policies, business partnerships, and priorities can change at any time.
What Cloud AI Actually Does With Your Data
Most people don't read the terms of service. Here's what they say:
Your conversations are stored on the provider's servers. Depending on the service, they're retained for 30 days, 18 months, or indefinitely. Some services use your conversations to train the next version of their model unless you find the opt-out toggle buried in settings. Human reviewers may read your chats for quality and safety evaluation. Metadata like your IP address, device information, timestamps, and usage patterns gets collected alongside the conversations themselves.
Even services that promise not to train on your data still retain it. Even services with strong privacy reputations still process your queries on their own infrastructure.
The question isn't whether you trust a specific company right now. It's whether you want your most personal thoughts to exist on infrastructure where the rules can change without asking you.
The Alternative: AI That Never Leaves Your iPhone
Your iPhone has a Neural Engine that processes 35 trillion operations per second on the A17 Pro. That's more than enough to run real AI models locally. No server. No cloud fallback. No "sometimes local, sometimes not."
Off Grid is a free, open-source app that runs AI entirely on your iPhone's hardware. After you download a model once, the app never contacts any server again. Airplane mode works perfectly.
There is no server. There is no account. There is no analytics, telemetry, or usage tracking. Not a single data packet leaves your phone. Not anonymous data, not personalized data. Nothing.
The code is on GitHub. MIT licensed. You don't have to trust a privacy policy. You can read every line yourself.
Onboarding
|
Text Generation
|
Image Generation
|
Vision
|
Attachments
|
What Runs on Your iPhone
Off Grid isn't a privacy demo. It runs six AI capabilities on your phone's silicon:
Text generation. Qwen 3, Llama 3.2, Gemma 3, Phi-4, or any GGUF model. Metal GPU acceleration and Neural Engine through Core ML. 15 to 30 tokens per second on A17 Pro and later.
Image generation. On-device Stable Diffusion through Apple's Core ML pipeline. 8 to 15 seconds per image on iPhone 15 Pro. 20+ models. Your prompts and images never leave your phone.
Document analysis. Attach PDFs, code, spreadsheets. Ask questions. Nothing gets uploaded anywhere.
Voice transcription. On-device Whisper speech to text. No audio ever touches a server.
Vision AI. Attach a photo or use your camera and ask questions. Processed entirely on device.
Tool calling. Models chain tools together in automatic loops. All running locally.
Why Architecture Matters More Than Promises
Every cloud AI company promises to protect your data. They all have privacy policies. They all say the right things.
But privacy policies are documents that can be rewritten. Companies make new partnerships. Business models evolve. What's true about a company's data practices today might not be true in six months.
When AI runs on your phone, promises don't matter. The model is a file on your device. The computation happens on your processor. Your conversations are bytes in local storage that only you can access.
There's no server to get breached. No database sitting somewhere with your conversation logs. No policy that can be updated to expand what's done with your data. No company that can be acquired by someone with different ideas about privacy.
It's not about trusting or distrusting anyone. It's about not needing to trust anyone at all.
Think About What You've Actually Typed
Health symptoms you haven't told anyone about. Questions about medications or conditions. Relationship problems you were processing before talking to the other person. Financial stress. Legal questions that reveal your situation. Drafts of difficult conversations. Therapy prompts. Personal journal entries.
Now think about the fact that all of those words are stored on a server you've never seen, managed by people you've never met, under policies you've never read.
With Off Grid, those conversations exist only on your phone. Not because a company promises not to look. Because there is nowhere else for them to exist.
Getting Started
- Install Off Grid from the App Store
- Download a model over WiFi (80MB to 4GB depending on what you need)
- Turn on airplane mode to verify the privacy claims yourself
- Use AI without your conversations living on someone else's server
Free. Open source. No subscription, no account, no ads. Your phone, your AI, your data.
Source code: GitHub
Top comments (0)