# Trust and privacy

> Who runs Artavolo.com, where the data lives (OVHcloud in Germany, or your own server), who else handles it, how it is protected, AI and your data, GDPR rights and the paperwork.

Source: https://artavolo.com/trust

What a security or privacy review asks for, in one place: who runs Artavolo, where your data lives, who else handles it, and how you take it with you. Everything on this page is how the service works today.

## The short answers

### Who runs Artavolo

jQWidgets Ltd, a company in Bulgaria, in the European Union. The people behind it have built data grids for developers for twenty years.

### Where your data lives

The hosted service runs on OVHcloud servers in Germany. Or run the same software on your own server, and nothing leaves it.

### Who can see your records

Only the people you give access to. Roles per workspace and base, which tables each person sees, and rules on which records, all checked on the server.

### What is recorded

An append-only audit log: who changed what, from what to what, when, and through which app or agent.

### Getting your data out

Export a table whenever you like, keep a backup of a base, and use the API. Leaving is as easy as arriving.

### Card details

Paid plans are paid through Stripe. Card numbers never reach our servers.

## Who else handles your data

The companies that process customer data for the hosted service. We announce a new one here 30 days before it starts.

| Company | What for | Where |
| --- | --- | --- |
| OVHcloud | Hosting: the servers, databases and files of the hosted service | Germany |
| Stripe | Payments for paid plans. Card details are typed into Stripe’s own page and never reach Artavolo. | Ireland (EU) and the United States |

### Only when you switch them on

These receive data only after someone in your workspace sets them up, and only what that setting sends.

| Service | What for | Where |
| --- | --- | --- |
| Mistral AI, OpenAI or Anthropic (Claude) | AI features, with the key and the service your workspace owner picks | France (EU) for Mistral AI; the United States for OpenAI and Anthropic |
| Slack and Twilio | Messages and texts your automations send | Their own regions |
| Google Drive, Dropbox, OneDrive | Files you choose to import | Their own regions |
| Your webhooks and your apps | Changes sent to addresses you set up, and Zapier, Make or n8n flows you connect | Wherever you run them |

## How your data is protected

The details, step by step, are in the [security overview](https://artavolo.com/docs/security) and the [architecture](https://artavolo.com/docs/architecture) document.

### Every request is checked

Before any data is read, the server checks the person against the workspace, base, table and record. That covers the browser, the APIs, AI chat, file downloads, live updates, exports, guests and forms.

### Encrypted on the way

Every connection to the service uses HTTPS.

### Sign-in

Two-factor authentication with one-time recovery codes for everyone; single sign-on with OpenID Connect (Entra ID, Okta, Keycloak) with the Enterprise licence. Passwords are stored as Argon2id hashes, never as text.

### Files

Kept in a folder per workspace and handed out only through the same permission check as records.

### API tokens and agents

An API token reaches one workspace, and only the bases named for it. An AI agent has its own identity, its own scope and its own line in the audit log.

### Deletes can be undone

Deleted records and tables go to a trash for 30 days, whoever or whatever deleted them.

### Tested

The boundaries between workspaces, people, tokens and guests are covered by automated test suites, run before each release.

## AI and your data

### Agents ask first

An AI agent writes as itself, never as you. Its changes can wait for a person to accept them, and an agent’s batch can be reverted from the audit log.

### AI chat acts as you

Claude, ChatGPT or Le Chat connected to Artavolo work with your permissions, only on the bases you allowed when connecting. Revoke a connection at any time.

### Your choice of AI service

AI features use the service your workspace owner sets up: Mistral AI in the EU, OpenAI, Claude (Anthropic), or a model inside your own network on your own server. Nothing goes to an AI service before that.

### No training on your records

Artavolo does not train AI models, on your records or on anything else.

[How AI agents work](https://artavolo.com/docs/agents) · [Connecting AI chat](https://artavolo.com/docs/ai-connector)

## Your rights, and the paperwork

### Data processing agreement

For business customers we sign a data processing agreement under Article 28 of the GDPR, with its list of technical and organisational measures and the sub-processors above. Ask for it through the [contact page](https://artavolo.com/contacts).

The privacy policy for the website and your account is [here](https://artavolo.com/page/privacy-policy).

### Access, export and erasure

Export your tables, keep backups of your bases and read everything through the API. Anyone can delete their own account from their profile; that removes their account details, sign-in links, tokens, permissions, history, notes and picture, and is recorded in the audit log.

Deleted records wait in the trash for 30 days, then are gone.

### Found a security problem?

Tell us through the [contact page](https://artavolo.com/contacts), with what you saw and how to see it again. We answer within two business days, and fix confirmed serious problems first.

### Need it on your own server?

The same software runs on your servers, with Docker or a plain PHP server: the Community edition is free, and the Enterprise licence adds single sign-on and scheduled encrypted backups. [Security and Enterprise](https://artavolo.com/enterprise).
