Security and privacy
What Assure keeps, where it lives, who can see it, and when it is destroyed. Each statement here matches how the system actually works.
- No test without permission
- A run cannot start unless the bot has an approved, signed permission letter whose dates cover today. The database refuses it otherwise, not just the screen.
- Test environments only
- Targets are test or staging environments. Production is not an option in the system.
- What we store
- Our synthetic personas, the messages our attacker agent sent, and your bot's replies. Personal data in a reply, such as a phone number or account number, is masked when it is stored: its type and position are kept, its value is not.
- The unmasked text
- The original reply is kept encrypted, readable only by our analysts in a two-step sign-in session, and every read is logged. It is destroyed 90 days after the report is approved, or sooner if your permission letter says so.
- Our own messages
- Anything that looks like personal data in our attacker agent's messages is masked for good. Those numbers are made up, so no original is kept.
- Model training
- Conversations are processed by a commercial model API whose terms exclude training on our data. Model keys are kept in an encrypted vault, never in code.
- Where data lives
- The database runs in the EU (Ireland). Test workers run on Oracle Cloud in Johannesburg. Nightly backups are encrypted before they leave the server, stored in Johannesburg and deleted after 30 days.
- Who can see what
- Access is by role. Your viewers see confirmed findings and approved reports only. Our owners and analysts must use two-step sign-in, and decisions are written to an audit log.
Questions from your security or data protection team are welcome. Contact us.