Skip to content
Let ideas be heard and improvements last.
Gemy

Trust centre

A mechanism behind every security sentence

Data location, encrypted backups, tenant isolation, AI policy, identity and privacy tools are on this page; each can be verified in the code and the setup. We state plainly what we do and what we don't do yet; we claim no independent audit or certification.

  • In the cloud or on your own server
  • AI starts switched off
  • Install on your own server as an option

In short

The cloud hosting provider and country are not yet stated on this page; if you choose your own server, data stays with you. Companies are isolated with database row-level security, backups are AES-256-GCM encrypted, AI starts switched off, and names, e-mail addresses, phone numbers and similar details are hidden in text sent to it (limits below). We claim no independent audit or certification; an internal pre-assessment was done, an external test is planned.

  • Option to install on your own server
  • Tenant isolation with RLS at the database level
  • AES-256-GCM encrypted backups, restore drills on a schedule
  • AI starts switched off, masking cannot be disabled
  • No ISO 27001 / SOC 2 certificate or independent penetration test yet

Trust

Your company's data, protected by mechanisms

Every sentence rests on a mechanism you can verify. Details are on this page; no independent audit or certification is claimed.

If you want data to stay in your own data centre, you can install on your own server.

  • Tenant isolation in the database

    Every row carries the company id; row-level security is forced (RLS + FORCE). Row-level security checks the company id on every query; isolation between companies is tested automatically in every release.

  • Evidence chain

    Critical approvals are kept in keyed, SHA-256-linked links; tampering is detected.

  • Health data vault

    Health documents live in sealed rows; only the company's authorized health role opens them, approvers see only "valid / invalid".

  • Encrypted backups and drills

    Backups are AES-256-GCM encrypted with a signed integrity list; restore drills run on a schedule.

  • Sign-in security

    PINs are one-way, salted and peppered with a server-side secret; two-step verification for e-mail sign-in and phishing-resistant passkeys for administrators; OIDC and SCIM are switched on by the company admin.

  • Browser and request protection

    Per-request nonce content security policy, HSTS, request and body limits; file scanning enabled at install.

Data location, backups and encryption

Data location

The cloud hosting provider and country are not yet stated here. If data must stay in a given country, install on your own server; data stays with you.

Encrypted backups

Backups are AES-256-GCM encrypted with a signed integrity list, and restore drills run on a schedule.

Data-loss target (RPO)

Continuous archiving can be switched on at install; the target is set on the admin screen. If archiving is off, loss runs to the last backup.

Encryption

Connections use TLS and HSTS; AI and integration keys entered on screen are stored encrypted; PINs are one-way, salted and peppered with a server-side secret.

Tenant isolation

Every row carries the company id and database row-level security is forced; isolation between companies is checked by automated tests in every release.

File safety

Uploaded files are checked for content and extension; virus scanning is switched on at install; files are reached only with a session and a permission check.

Audit trail

Status changes and critical operations are written to append-only records; critical approvals are linked in an evidence chain.

AI policy

AI is an option, not a requirement. Your company decides.

Off by default

AI starts switched off in a new company; the company admin switches it on, or off again at any time.

Sending out is the company's decision

No text is sent to a cloud AI service until a company officer allows it; provider, country and contract details are asked for. A model on your own server can be chosen instead.

Default: services that don't train on your data

We don't use customer data to train models. A new company can only choose services that don't use data for training; if the company switches this level to "all", a service on a free plan may use the data for training.

Masking

In text sent to AI, the names of people in the company, common first names, staff numbers (a numbers-only staff number only after the word “sicil”), e-mail, phone, national ID number, IBAN and plates are hidden; this step cannot be switched off. A surname on its own and other personal details typed into free text (such as an address or a health condition) are not hidden.

People decide

AI suggests; a person approves before anything is applied. No record, score or risk class is written automatically.

Outside connections start off

None of the features that connect outside, such as API keys, webhooks, corporate sign-in and AI, work until the company switches them on.

Identity and access

Staff number and PIN, no e-mail needed

An employee without e-mail signs in with a staff number and a 6-digit PIN; lockout grows after wrong attempts, and sessions on a shared tablet are short.

  • PINs are one-way, salted and peppered with a server-side secret
  • PIN cards are printed as PDF from the admin panel
  • Shared-tablet (kiosk) mode runs with short sessions
Sample company · demo dataSign-in with staff number and PIN
Sign-in with staff number and PIN

Verification codes, passkeys and corporate sign-in

A verification code for e-mail sign-in; administrators can use phishing-resistant passkeys. Roles carry the responsibilities used in approval flows and decide who sees what.

  • OIDC sign-in with your own identity provider; switched on by the company admin
  • SCIM user sync; switched on by the company admin
  • Session and object permission are checked on every request
Sample company · demo dataRoles and permissions: system roles and the company's own roles
Roles and permissions

Checked in every release

Three layers: self-assessment, an internal pre-assessment and automated scans. An independent penetration test is planned.

  1. 01

    OWASP ASVS 5.0 Level 2 self-assessment

    Status, evidence (tests) and open work are written for each chapter; some chapters are "partly" met. This is a self-assessment, not a certificate.

  2. 02

    Internal pre-assessment

    An attacker-minded internal review in October 2026: no critical findings; all high and medium findings were closed with regression tests.

  3. 03

    Scans in every release

    Code scanning, dependency audit, image scanning and cross-company isolation tests run in every release.

Personal data and privacy tools

Your company is the data controller; we act as processor on the company's instructions.

Export

A company admin can download all company data or a single person's data.

Leavers

A leaver's personal data is anonymised, respecting legal retention periods.

Deletion request

A company deletion request waits 30 days; it can be cancelled in that time, then becomes permanent.

Support access

Support sees only a configuration summary, with the admin's time-limited approval (up to 8 hours); no record content or personal data; every access is logged.

AI

The company switches it on or off; outgoing text is masked by rules, with the limits stated on the AI page; applying a suggestion needs human approval.

Health data

Kept in a separate vault; only the health role authorised by the company can open it.

Known limits

We list what we don't have yet too, so you can decide knowing it.

No certification

No independent penetration test or ISO 27001 / SOC 2 certificate yet; an internal pre-assessment was done, an external test is planned.

Needs setup

OIDC single sign-on and SCIM user sync are not on by default; the company admin switches them on and connects the company's identity provider.

No ready-made ERP connector

There is no ready-made connection for SAP, Logo or Netsis; ERP and MES are integrated through the read-only API, webhooks and registered Excel imports, all switched on by the company.

Hosting

The cloud hosting provider and country are not yet stated on this page; if you have a data-location requirement, choose installation on your own server.

Security questions

Can other companies see our data?

No. Companies are isolated with PostgreSQL row-level security, verified by automated tests in every release.

What do you offer for privacy compliance?

Full ZIP export, anonymisation of leavers, a 30-day deletion request, privacy notice acknowledgement and an audit trail.

Cookies

Essential cookies are needed for the site to work. Analytics cookies load only if you allow them; declining is as easy as accepting. You can change your choice at any time via the footer link.

Cookie policyPrivacy notice