The Russian software developer FIS has introduced a new service designed to anonymize corporate documents before they are processed by generative AI models. This solution allows organizations to strip sensitive information locally while maintaining the analytical context required for tasks such as document comparison, summarization, and contradiction detection.
The service identifies and replaces sensitive identifiers like names, company titles, and financial details with placeholders. It supports both reversible masking, which allows for the restoration of original data via a secure internal mapping table, and irreversible deletion for non-essential information. Users can also define custom replacement rules.
The platform is compatible with various document formats, including PDF, DOCX, XLSX, PPTX, and VSDX. It also features text recognition capabilities for images and diagrams. The system can be deployed within a customer's own infrastructure or a secure cloud environment, offering access via a web interface, API, or development environments.
Security features include role-based access control, operation logging, and the isolated storage of documents and mapping tables. By processing sensitive data internally, the service enables banks and other organizations to leverage external generative AI models while adhering to strict information security policies.
