Crowlin / AI and data protection
Company AI and data protection.
Contracts, salaries, prices and personal data don’t belong in a public chatbot. Crowlin is designed so that company documents stay in the environment you choose and everyone sees only what they may.
Questions your lawyer and IT will ask.
Before a company lets AI near internal documents, it needs answers to five things.
Where does the data live?
In a secure environment in the EU (a data centre in Germany), on a dedicated server, or on your own infrastructure. You choose.
Who processes it?
Answers from documents are generated by a language model in that environment. We don’t send documents to external AI providers when processing document questions.
Who can access it?
Only signed-in employees, by role, team and confidentiality level. The server checks this before every search.
Can we trace who saw what?
Yes. Every question is logged: who asked, when, what they were allowed to see and which documents the answer came from.
What leaves?
Only what you explicitly allow: a web search query (can be switched off), or tasks of digital employees for which you choose a cloud model.
What stays and what leaves.
We document the data flow for the chosen setup when designing the solution, so you can record it in your processing records.
Stays in your environment
Your documents, their content and the answers drawn from them. Question history and audit records. Only encrypted backups are stored off the server.
Leaves only with your permission
A query to the web search engine (optional, can be switched off — only the search term goes out). Tasks of digital employees, if you choose a cloud model for them.
GDPR and contracts.
Crowlin won’t make you GDPR compliant on its own. It gives you the tools you need to get there.
Data processing agreement
When we host Crowlin, we sign a data processing agreement and document the data flow for the chosen setup.
Only the sources you want
You connect only what Crowlin needs. We delete demo or pilot data on request.
Role-based access
Permissions come from roles and teams. Changes apply immediately, including previously cached answers.
Common questions about data protection
Is the model trained on our documents?
Can an employee reach salary data with a clever question?
Where are the backups?
Can we run Crowlin entirely on our premises?
Related topics
We’ll show you on your own data.
We’ll show you Crowlin on a sample of your documents in an isolated space. We delete the data after the demo on request. The demo is non-binding.
How the first demo works
- Intro call, 30 minutes. What people in your company ask about and which documents hold the answers.
- Demo on your documents, about an hour. Online or on site. You send a sample; Crowlin answers from it live.
- Pilot proposal. Scope, success criteria, timeline and a fixed price. You decide only then.