micapa.ai

micapa.ai is a project built by Sozpic to automate the detection and redaction of sensitive information in documents and texts.

The system combines a specialized NER model with a multi-stage processing pipeline, trained and validated on more than 80,000 documents, to identify personal and confidential data more accurately and to cope with real documents, with very different structures, formats and contexts.

It can detect information such as names, ID documents, addresses, phone numbers, email addresses, companies and any other entity that could identify a person or expose private information.

Unlike systems based only on rules or regular expressions, micapa.ai analyzes the context in which each piece of data appears and combines several detection techniques. That lets it handle complex documentation and cut errors both from missed data and from hiding information that didn’t need hiding.

The goal is simple: turn real documents into information that can be used, processed or analyzed while keeping the exposure of sensitive data to a minimum.

The project is aimed above all at organizations that work with large volumes of documents, such as professional firms, insurers, companies, public administrations or teams that need to use internal information in artificial intelligence processes.

micapa.ai combines AI, natural language processing, NER models and sensitive-data protection systems to add a layer of security before the information is processed.

This is only the first phase of the project. In a second stage, micapa.ai’s technology will be built directly into different applications and workflows, so sensitive information can be detected and hidden automatically wherever it is created, consulted or processed.

A Sozpic project that tackles one of the key needs of the new era of artificial intelligence: being able to use data without exposing the people behind it.