Security & trust
What we keep, and what we don’t.
Each RAVIONIX tool keeps only what it needs to work, and this page says what that is for the tools that are live today. Some of it you can check yourself: watch the network traffic and you'll see a password never leaves your browser, only its hash. The rest describes how our servers are built.
RAVIONIX is not currently SOC 2 or ISO 27001 certified, and no completed third-party attestation is implied. No controls have been independently assessed and no penetration test has been done yet. The Trust Center documents the current controls and certification roadmap.
Raw content is processed, not stored
Links and messages are processed in memory to produce a verdict and are not stored in the RAVIONIX database. Our AI provider reads them to judge them. We keep a one-way hash, the verdict, the confidence, the AI's short explanation and basic details such as the time. If a check is rated a scam with high confidence, we also keep the website's domain name, never linked to you.
Passwords never leave the device
Your browser hashes the password; we receive only the hash and send Have I Been Pwned just its first 5 characters. We keep that 5-character prefix and the result.
Locked audit log
A database trigger blocks edits and deletes. Entries are removed by a scheduled job after a set period. Deleting your account removes your account ID from them.
Export and delete your data
Account holders can export or delete their data from settings, without contacting support.
How it’s built today
How the live tools work right now.
Raw content is not stored
Processed to produce a verdict, never stored by us (our AI provider keeps API data only for a limited time). We keep a hash, the verdict and the AI's explanation, plus the site name of links rated a scam with high confidence, never linked to you
Passwords never leave the device
Hashed in the browser, checked by k-anonymity
Every AI answer checked
Each answer must match a fixed format before you see it. A repeat of a recent check can reuse its verdict.
Export and delete your data
Account holders can do it from settings, without contacting support
WCAG 2.2 AA target
We work toward WCAG 2.2 AA and fix accessibility issues as we find them. Not independently audited yet.
Everything above describes how RAVIONIX Scanner works today.
Scam sites we collect, to help protect others
When our AI rates a link a scam with high confidence, we save the website’s name so a person can review it. Confirmed scam sites can then help warn other people. Here is how it works. The exact time limits are on our Privacy page.
1. Flagged
We save only the site’s name (for example secure.paypa1-login.com), how many times it was flagged, and the day. Never the full link, who checked it, their account or their IP address.
2. Reviewed by a person
Once a site has been flagged several times, a member of our team looks at it. The AI alone never decides.
3. Confirmed or cleared
Confirmed scam sites are marked as scams and may be shared in a future public blocklist. Sites that turn out to be safe are cleared. Every decision is logged with the date and a note.
4. Cleaned up
Site names nobody has reviewed are deleted after a set period. Reviewed sites keep their history, so a safe site isn’t flagged as new again.
Is your site flagged by mistake?
If you own a site we have marked as a scam, or it has been cleaned up since, email us with the site name. A person will look at it again. If it is safe, we clear it and note why. We keep the dates it was flagged and reviewed.
In development
RAVIONIX AI Guard
AI Guard isn’t released yet, so these are design goals, not tested claims.
About AI Guard- Designed to work offline, with no uploads and no telemetry.
- Designed never to read your AI conversations or chat history.
- Designed never to show or save a full secret.
- Designed never to run the tools it inspects.