How we use AI
- Document validation (passport scan machine-readable zone (MRZ) validity, photo specifications: size, background, composition, expired document detection)
- Application completeness (flagging missing or inconsistent fields before human review)
- Information extraction (passport number, expiry date, nationality)
Systems are assistive and do not make final decisions.
Human review
Every application is reviewed by a human team member before any action is taken; no application is submitted to a government authority based solely on an automated decision, and a person signs off on the documents.
What the automated checks do, and what they do not
Checks flag readable passport data, whether a single face can be detected in a photo and whether the upload is large enough to be used, and completeness. They do not certify that a photograph meets a government's requirements: head size, tilt, eye line, shadows and background are not measured automatically. Where a check cannot confirm something, it is passed to human review rather than treated as a pass.
Privacy safeguards
Document data sent to AI validation systems is transmitted over encrypted channels and is not used to train or improve third-party AI models; passport and photo files are held in private storage with time-limited access links and the copies used for automated checks are removed once the check has run; documents are not shared with external AI service providers for purposes other than the specific validation task.
Transparency
Applicants are told what was detected when an application is flagged and what needs correcting; they can ask whether a particular decision involved automated processing via support@haavisa.com.
Changes to this policy
Updated as AI use evolves; material changes affecting data processing are notified by email before taking effect.