Roborana applies GDPR principles from day one: data minimisation, purpose limitation, and storage restriction
Data minimisation: we only use the data the AI step actually needs. If invoice processing requires PO number, amount, and date, we don't feed in supplier phone numbers or employee details.
No training on your data by default. Off-the-shelf models (OpenAI, Azure AI) are pre-trained on public data, not your client data. When custom models are necessary, we work with anonymised or synthetic data where possible and require explicit permission before using real personal data.
EU data residency: Azure AI services process data within EU data centres. Data does not leave the EU unless you explicitly authorise it. For Belgian organisations with strict GDPR requirements, this is essential.
Access control and encryption: role-based access restricts who can view and modify data. All data is encrypted in transit and at rest. Unauthorised access triggers alerts. Data retention policies ensure information is deleted once it's no longer needed.
Privacy by design: we consider privacy at every decision point. Can we delete the original document after extraction? Do we need detailed logs or just summaries? Each choice reduces exposure incrementally.



Send us a message...