Anonymization of Sensitive Information in Financial Documents Using Python

Learn to anonymize sensitive information in financial documents using Python, diffusion models, and named entity recognition in this 27-minute conference talk from EuroPython 2025. Discover how to leverage open-source tools and self-hosted models to replace personally identifiable information (PII) with realistic synthetic alternatives while preserving document integrity. Explore various approaches for Named Entity Recognition (NER) to identify sensitive entities and understand how diffusion models can inpaint anonymized content. Address the critical challenge faced by financial, medical, and legal institutions that struggle to utilize their own data for machine learning due to privacy, compliance, and security requirements. Master practical techniques for creating high-quality training datasets while maintaining strict data protection standards, enabling organizations to develop better models without compromising sensitive information.