TL;DR
We collect nothing. Zero. Zilch. Atlas Cortex runs entirely on your device. There are no servers, no accounts, no analytics, and no telemetry. Your data never leaves your hardware.
1. What Atlas Cortex Is
Atlas Cortex is open-source software that runs locally on your personal hardware. It is not a hosted service, cloud product, or SaaS platform. When you use Atlas Cortex, you are running the software on your own device — we have no access to your device, your data, or your usage.
2. Data We Collect
None.
Atlas Cortex does not:
- Collect personal information
- Track usage or behavior
- Send telemetry or analytics
- Phone home to any server
- Require an account or registration
- Use cookies, fingerprinting, or any tracking technology
3. Data Stored on Your Device
Atlas Cortex stores data locally on your device to function. This includes:
- Conversations: Chat history is stored in a local SQLite database.
- Memory: Long-term knowledge and context is stored locally.
- Configuration: Your settings, preferences, and Home Assistant tokens are stored in local files.
- Models: AI model weights are downloaded from Hugging Face and cached locally.
- LoRA Adapters: Fine-tuned adapters are stored in local directories.
All of this data is under your full control. You can inspect it, back it up, move it, or delete it at any time.
4. Camera, Microphone & Sensors
Atlas Sight and voice features may access your device's camera and microphone. All processing happens entirely on your local hardware:
- Camera: Video frames are processed by on-device vision models. No images or video are transmitted, uploaded, or stored remotely.
- Microphone: Audio is transcribed locally using on-device speech-to-text models (Whisper.cpp). No audio recordings leave your device.
- Text-to-Speech: Voice responses are generated locally. No text is sent to cloud services for synthesis.
You can verify this by inspecting the source code or monitoring your network traffic — Atlas Cortex makes zero outbound connections for any media processing.
5. Third-Party Services
Atlas Cortex is designed to be 100% offline. However, some optional features may connect to external services only when you explicitly configure them:
- Home Assistant: If you configure a Home Assistant URL and token, Atlas Cortex will communicate with your Home Assistant instance on your local network.
- Weather/News APIs: Some plugins may fetch data from public APIs if you enable them. These requests do not include personal data.
- Hugging Face: Model downloads come from Hugging Face. Standard Hugging Face download policies apply.
None of these connections send your conversations, personal data, or usage patterns to any third party.
6. Children's Privacy
Atlas Little Learner is designed for children ages 18 months to 5 years. Because all processing happens on-device, no children's data is collected, stored remotely, or shared with anyone. The application includes content safety guardrails and age-appropriate content filtering.
7. This Website
This website (atlascortex.dev) is a static site hosted on GitHub Pages. We do not use cookies, analytics, or tracking scripts. GitHub Pages may collect standard server logs (IP addresses, user agents) as part of their hosting service — see GitHub's privacy practices for details.
8. Open Source Verification
Atlas Cortex is released under the MIT License. The entire source code is publicly available on GitHub. You can audit every line of code to verify these claims yourself.
9. GDPR, CCPA & Regulatory Compliance
Atlas Cortex is compliant with GDPR, CCPA, and similar data protection regulations by design. Because no personal data is collected, transmitted, or processed by us, there is nothing to regulate:
- Right to access: There is no data held by us to access.
- Right to deletion: All data is on your device — delete it anytime.
- Right to portability: Your local SQLite databases and config files are yours to move freely.
- Data processing agreements: Not applicable — we never process your data.
- Data breach notification: Not applicable — we hold no data to breach.
No data means no compliance burden. This is privacy by architecture, not by policy.
10. Changes to This Policy
If we ever change this policy, it will be to collect even less data (which is hard when the answer is already zero). Any changes will be reflected in the "Last updated" date above and committed to the public repository.
11. Contact
Questions about this policy? Open an issue on GitHub.