How we protect the Arabic data you send us, and the questions to ask any Arabic data vendor, including us.
Arabic training and evaluation data is often personal, sensitive or regulated: customer calls, service chats, clinical notes, banking conversations. Here is how we handle it by default, and the questions you should ask any Arabic data vendor, including us.
| Hosting | In-region hosting by default, with on-prem and private-cloud options for mission-critical work. |
| Access | Least-privilege access: contributors see only the items their task needs. |
| Audit | Audit logs and an audit trail on every task. |
| People | Vetted contributors under NDA. Everyone passes dialect and domain screening before touching client data. |
| Quality | Gold-standard calibration, adjudication and rework loops, so errors are caught before delivery. |
| Independence | We do data only. We're model-agnostic and never compete with our clients' models. |
Use these in your security review. Ask us too; we'd rather answer them up front than after the contract.
If your data includes personal data covered by Saudi Arabia's Personal Data Protection Law, read our guide to PDPL, data residency and AI training data (also in Arabic). It covers cross-border transfer conditions and what they mean for annotation and evaluation work.
Include your security questionnaire or data-protection requirements with your pilot scoping request, or email hello@bayanatlabs.com. See also how a pilot works.