The security file
How your data is handled,[1][1] Without the badge wall: where a control does not exist, it is not listed here. Where we hold no certification, chapter 04 says so rather than letting a procurement form assume one. without the badge wall
This page describes controls that exist in the product today. Where a control does not exist, it is not listed. Where we hold no certification, we say so rather than leaving you to assume.
Your prompts and your responses are your data. Everything below is specific enough to check, and nothing below is aspirational.
01: Authentication
Sessions your browser cannot leak
Most token theft in single-page applications comes down to one decision: whether a token was ever put somewhere JavaScript could read it. Ours is not.
Tokens never reach the browser
Sign-in runs through Zitadel OIDC behind a backend-for-frontend. Access and refresh tokens are held server-side. Nothing that could be replayed is exposed to page JavaScript.
Sessions are cookies, not storage
The session cookie is httpOnly and SameSite=Lax. It is not in localStorage, which means a script injected into the page cannot read it.
TLS everywhere it is deployed
Every deployed environment serves over TLS, and session cookies carry the Secure flag.
02: Access control
Who can reach what, and what gets recorded
Isolation and roles are enforced in the API. A control that only hides a button in the interface is not a control.
Workspace and brand isolation
A workspace is a hard boundary. Its prompts, responses, competitors and evidence are not reachable from another workspace, including for members of both.
Two roles, enforced server-side
Admin can write. Everyone else is read-only. That is checked on the server, not hidden in the interface, so a read-only member cannot write by calling the API directly.
Revocable API keys
Keys are stored as SHA-256 hashes, never in plaintext. Revoking one takes effect immediately. The key value is shown once, at creation.
Audit logging on administrative actions
Administrative actions are written to an append-only log. Being precise about scope: this is an admin-side log, not a per-workspace audit trail you can browse. If you need the latter, it does not exist yet.
03: Your data
What we hold, and what leaves
Measuring what a model says about you requires sending it a question. That is the one place your data leaves our systems.
What we store
The prompts you track, the responses we measured, the sources those responses cited, and the scores computed from them. Findings stay auditable because the underlying response is still there.
What goes to providers
Your prompts, to the model providers listed on the engines page, in order to run them. That is the whole outbound surface. Your prompts and responses are not used to train models.
See which providersYour rights
Deletion and export are handled through the process in our privacy policy. For a specific compliance question, ask us directly rather than working it out from this page.
Read the privacy policy04: certifications
We hold none
Not SOC 2. Not ISO 27001. Not HIPAA. No audit is in progress that we can point you at.
We could describe the controls above as enterprise-grade and let a procurement form fill in the rest. We would rather lose a checkbox now than have you discover this in week three of an evaluation. If a certification is a hard requirement, this is the paragraph that saves us both a fortnight.
Send us a security questionnaire05: Questions
Security, answered plainly
Are you SOC 2 or ISO 27001 certified?
No. We hold no security certifications and have no audit in progress. If that is a hard requirement, we are not a fit yet, and we would rather tell you now.
Where are my authentication tokens stored?
Server-side, behind a backend-for-frontend. The browser only ever holds an httpOnly, SameSite=Lax session cookie. Tokens are never exposed to page JavaScript and are not in localStorage.
What happens to the prompts and answers you collect?
They are stored so findings can be opened and checked. Prompts are sent to the model providers listed on our engines page in order to run them. Neither prompts nor responses are used to train models.
Can I control what my team can see and change?
Yes, within two roles. Admin can write, everyone else is read-only, enforced server-side. Access is granted per workspace. A finer permission model does not exist yet.
How do I delete my data?
Follow the process in our privacy policy, or contact us and we will handle it. Deleting a workspace removes its prompts, responses and evidence.
- “Without the badge wall” describes the whole page, not just chapter 04. Every card above states what is enforced and where the boundary sits, rather than reaching for a word like “enterprise-grade” that a reader cannot check.
Ask us the awkward questions.
The ones a security review would ask on page four. We would rather answer them now.