← Back
Privacy Policy
Effective: 2026-07-13 · Extension: Intelligence Engine (Chrome, Manifest V3) · Publisher: Emrj Technologies · hussein@nextdecentrum.com
The short version
Intelligence Engine runs entirely on your device. There is no server. There is no account. The only network requests it ever makes are the one-time downloads of the AI models it needs to run — those downloads carry no user data.
What we collect
Nothing on our servers. We have no servers.
What the extension stores on your device (IndexedDB):
- Anonymized reasoning segments extracted from your AI conversations. Segments where an identified entity survives verbatim after anonymization are dropped, not stored.
- A 384-dimensional embedding vector per segment so search works.
- Metadata: domain, category, platform (Claude / ChatGPT / Perplexity / Gemini), timestamps.
What the extension stores in chrome.storage.local:
- Capture preferences (global toggle, per-platform toggles, per-conversation pauses).
- Model load status.
- Installation timestamp and version.
What we do not collect
- Raw prompts or AI responses. They are processed in-memory inside the extension's offscreen document and discarded after anonymization.
- Names, companies, locations, monetary amounts, or dates. An on-device NER model detects these and an on-device LLM (or deterministic replacer) rewrites the segment to remove them before anything is stored.
- URLs, browsing history, or clicks. The extension only sees URLs on the six declared AI-tool hosts, and only to track per-conversation pause state.
- Analytics, telemetry, error reporting, or any beacons. None.
- Your identity in any form. No account, no login, no cookies set by the extension.
Network requests the extension makes
On first use, the extension downloads its AI model weights from three public content-delivery hosts:
- huggingface.co — Transformers.js models (all-MiniLM-L6-v2, bert-base-NER) and MLC shards
- cdn.jsdelivr.net — Transformers.js runtime shards
- raw.githubusercontent.com — MLC / WebLLM catalog files
These requests are software fetching itself. They carry no user identifiers, no captured conversation content, no cookies. Once downloaded, models are cached by the browser and re-used offline. After the initial download the extension makes no further network requests during normal operation.
This landing page
This site is behind a password so we can share early. If you sign up for the waitlist, we store your email address (only) in a single database row so we can send you an invite. That row is deleted when we invite you. No cookies, no tracking, no analytics on this page.
Where your data lives
On your device. In this browser's IndexedDB and chrome.storage.local. You can export it any time as Markdown, JSON, or HTML. You can delete it any time — per-entry, per-domain, or entirely. Uninstalling the extension deletes all associated storage per Chrome's extension lifecycle.
Sharing
We do not share your data. We cannot — it never leaves your device.
Children
Intelligence Engine is intended for professional use by adults. It is not directed at children under 13.
Changes
If we change this policy, the updated version will appear here. The effective date at the top will change.
Contact
Questions or requests: hussein@nextdecentrum.com
The full extension source, including the anonymization pipeline, is available for inspection. Ask.