How RAVIONIX is built.

What we store, and what we do not

The links, messages and emails you check are hashed (SHA-256) at our Cloudflare edge. The text is passed in memory to our API and to our AI provider, Anthropic, to produce a verdict. We never store it (Anthropic may keep it for up to 30 days for safety review and does not train on it). We keep only a one-way hash, the verdict and the AI's reasoning. If a check is rated a scam with 80%+ confidence, we also keep the website's domain name — never the full link, never linked to you. Domains confirmed as scams after human review (at 5 or more flags) may be published in a public blocklist used by RAVIONIX Guard; unconfirmed names are deleted 90 days after they were last flagged. Passwords are hashed in your browser and never leave your device; we keep only a 5-character prefix of that hash.

Data protection

• All traffic is encrypted in transit with TLS.
• Data at rest is encrypted by our database provider. Especially sensitive fields, such as monitored email addresses, are additionally encrypted by us with AES-256-GCM.
• Secrets live in our hosting provider's encrypted environment settings, never in source code.

Sign-in and access

• Sign-in is handled by Clerk with Google or Microsoft accounts, so your Google or Microsoft 2-step verification, if you've turned it on, also protects your RAVIONIX sign-in. RAVIONIX never stores a password.
• Every request is checked against the account it belongs to, so one customer can't reach another's data.
• When Teams launch, they will have owner, admin and member roles, with billing and member changes limited to the right roles.
• Internal admin access is restricted to a named allowlist, and every admin view is written to the audit log.

Abuse and AI safeguards

• Cloudflare Turnstile bot checks and server-side rate limits on every free check.
• We tell the AI to treat what you submit as data, not instructions, and we check every AI answer before it is shown.
• A monthly AI spending cap protects the service from runaway use.

How we work

• Automated tests, secret scanning and dependency checks run on every change.
• Security-relevant events go to an audit log that is locked against editing or deletion and kept for up to 2 years.
• Our database provider offers point-in-time recovery.
• No availability figure is published: uptime is not yet measured against a committed target.