How Our AI Works
Malin uses artificial intelligence to automatically understand documents, extract data, and suggest actions. Here we transparently explain how it works.
Which AI Models We Use
Malin uses Claude (Anthropic) and Gemini (Google) for document analysis and data extraction.
Our AI pipeline includes: OCR processing for text recognition, field extraction (amounts, dates, parties), automatic document classification, and confidence scoring for every extraction.
What Data is Processed
Document content (text and images) is sent to AI for analysis.
Extracted fields include: amounts, dates, involved parties, and categories.
Confidence Scores and Evidence
Every AI extraction includes a confidence score (0–100%).
Evidence linking: each extracted field links back to the source region in the document.
Low-confidence fields are automatically flagged for human review.
Human Oversight
No AI decision is executed without human approval.
Users can override any AI suggestion.
Full audit trail of all AI decisions and human actions.
Your Rights
- Right to explanation of any AI decision
- Right to contest AI-extracted data
- Right to request complete data export (GDPR)
- Right to account deletion
EU AI Act Compliance
Malin classifies as a limited-risk AI system (document processing, not biometric/social scoring).
Transparency obligations: model disclosure, confidence indicators, human oversight.
August 2026 enforcement timeline acknowledged.
We commit to continuous monitoring and adaptation.