Secure handling of Personally Identifiable Information (PII) in document exchange is critical with the advent of the digital age for data privacy as well as regulatory compliance. RedactSafe is an end-to-end framework that carries out automated detection, redaction, and secure storage of PII with the help of state-of-the-art Natural Language Processing methods like Optical Character Recognition (OCR) and Named Entity Recognition (NER). The system integrates rule based approaches with machine learning algorithms to identify sensitive data in both structured and unstructured documents effectively. Identified PII is then redacted and saved on the InterPlanetary File System (IPFS), while integrity and access history are recorded immutably using Ethereum smart contracts. A blended access control system, combining Attribute-Based Access Control (ABAC) and Role-Based Access Control (RBAC), restricts the ability of unauthorized users to view complete documents, with each request logged on-chain for auditing. In Experimental results, its is observed that there is high recall and precision in detecting PII, providing its effectiveness in maintaining documents confidential while remaining easy to use. The novelty of this work lies in its integrated approach-merging PII detection, redaction, audits logs and dynamic access. In sensitive environment, RedactSafe gives a scalable, privacy-protecting, and decentralized approach to secure document workflows.