Towards a HIPAA Compliant Agentic AI System in Healthcare CCS: Security and privacy Data anonymization and sanitization CCS: Social and professional topics Privacy policies CCS: Security and privacy Access control CCS: Computing methodologies Natural language processing

Subash Neupane, Sudip Mittal, Shahram Rahimi · 2025

Read Towards a HIPAA Compliant Agentic AI System in Healthcare CCS: Security and privacy Data anonymization and sanitization CCS: Social and professional topics Privacy policies CCS: Security and privacy Access control CCS: Computing methodologies Natural language processing by Subash Neupane, Sudip Mittal, Shahram Rahimi (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-17669.txt.

Introduction

Subash Neupane, Sudip Mittal, and Shahram Rahimi design an agentic health AI that respects United States privacy law in a 2025 study. Language model agents can draft reports and summarize cases on their own, working directly on protected health information. The health privacy law sets strict rules for such data, so autonomy must be fenced by anonymization, access control, and policy checks.

The authors present their design as a work in progress toward a compliant system. Getting compliance right early matters: a summarizing agent that leaks patient details would end adoption of far safer tools along with it.

Catalog record

CREATORSubash Neupane, Sudip Mittal, Shahram Rahimi
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE25 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-17669

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Towards a HIPAA Compliant Agentic AI System in Healthcare CCS: Security and privacy Data anonymization and sanitization CCS: Social and professional topics Privacy policies CCS: Security and privacy Access control CCS: Computing methodologies Natural language processing Subash Neupane, Sudip Mittal, Shahram Rahimi Abstract . Agentic AI systems powered by Large Language Models (LLMs) as their foundational reasoning engine, are transforming clinical workflows such as medical report generation and clinical summarization by autonomously analyzing sensitive healthcare data and executing decisions with minimal human oversight. However, their adoption demands strict compliance with regulatory frameworks such as Health Insurance Portability and Accountability Act (HIPAA), particularly when handling Protected Health Information (PHI). This work-in-progress paper introduces a HIPAA-compliant Agentic AI framework that enforces regulatory compliance through dynamic, context-aware policy enforcement. Our framework integrates three core mechanisms: (1) Attribute-Based Access Control (ABAC) for granular PHI governance, (2) a hybrid PHI sanitization pipeline combining regex patterns and BERT-based model to minimize leakage, and (3) immutable audit trails for compliance verification. 1. Introduction Agentic AI system is an emerging paradigm in Artificial Intelligence (AI) where autonomous systems pursue complex goals with minimal human intervention ( shavit2023practices ) . The integration of these systems powered by Large Language Models (LLMs) into healthcare workflows represents a paradigm shift in clinical decision making and administrative efficiency ( sudarshan2024agentic ) . Unlike passive LLMs, Agentic AI systems can dynamically interact with Electronic Health Records (EHRs), synthesize multimodal patient data, and execute context-aware tasks such as generating clinical documentation, summarizing medical literature, or offering real-time diagnostic recommendations ( sudarshan2024agentic ; 10.1145/3709365 ; 10825266 ) . Figure 1 . A graphical overview of the HIPAA-compliant Agentic framework. […] (used for brevity) indicates that there are more data modalities and downstream tasks. However, the autonomous nature of these systems introduces critical risks to the security of Protected Health Information (PHI) as mandated by the Health Insurance Portability and Accountability Act (HIPAA) ( act1996health ) . PHI is defined by the U.S. Department of Health and Human Services (HHS) in the HIPAA Privacy Rule (45 CFR § 160.103) as “ Individually identifiable health information ” that is transmitted or maintained in any form or medium (electronic, paper, or oral) by a covered entity or its business associates, and relates to past, present, or future physical or mental health or condition of an individual, provision of health care to an individual, and past, present, or future payment for the provision of health care to the individual . An example containing multiple PHI elements (name, birth date, medical record number, diagnosis, treatment information, visit date, and insurance identifier) that would require protection under HIPAA is shown in Fig. 2 . Patient Name: Barry Berkman DOB: 01/15/1965 Medical Record Number: MRN-12345678 Diagnosis: Type 2 Diabetes Medication: Metformin 500mg twice daily Last Visit: 03/02/2025 Insurance ID: ABC123456789 Figure 2 . An example of PHI identifiers that requires protection based on HIPAA Safe Harbor rule § 164.514(b)(2). HIPAA regulations require healthcare providers to implement technical, physical, and administrative safeguards to restrict access to PHI and ensure that only authorized users can interact with patient data. A critical component of these safeguards is access control, which governs who can view, modify, or transmit sensitive information. However, in the context of Electronic Health Records (EHR), where clinical notes (e.g. diagnoses, treatment plans, medical histories) are often stored as unstructured free text traditional access control mechanisms struggle to enforce granular protections. Unstructured data inherently contains sensitive identifiers (e.g. names, addresses) and complex clinical narratives, creating vulnerabilities that static access policies cannot fully mitigate. These risks are amplified with the integration of Agentic AI systems powered by LLMs in healthcare workflow. Without dynamic safeguards, LLM driven workflows risk inadvertently exposing PHI, memorizing sensitive details during training, or bypassing rigid access rules, violating HIPAA’s Minimum Necessary Standard (§ 164.502(b)). The current literature predominantly explores Agentic AI systems for specific downstream tasks, such as identifying cognitive concerns in clinical notes ( tian2025agentic ) , generating medical reports ( sudarshan2024agentic ) , and delivering clinical services ( reddy2025enabling ) . Nevertheless, these systems demonstrate significant limitations in achieving comprehensive HIPAA regulatory compliance, thus constraining their potential for widespread clinical deployment. To address regulatory and compliance gaps in previous works, this work-in-progress paper introduces a novel HIPAA compliant Agentic AI framework that provides technical safegurad to protect PHI. Fig. 1 provides a high-level overview of our framework for several downstream tasks such as diagnosis prediction, care plans, clinical summary generation, radiology report generation, and question and answer. Our framework implements HIPAA mandates through three core mechanisms: Attribute-Based Access Control (ABAC) (§ 164.312(a)(1)), which dynamically restricts data access based on user roles, resource sensitivity, and environmental context; layered PHI sanitization (§ 164.514(b)(2)), applying dual redaction stages (pre- and post-inference) to minimize PHI exposure; and immutable audit trails (§ 164.312(b)), which log all access events and policy decisions for compliance verification. The major contributions of this paper are: • An Agentic AI framework integrating ABAC, sanitization, and audit agents to enforce HIPAA compliance in healthcare workflows. • A hybrid PHI sanitization pipeline that combines regex patterns and BERT-based model to mitigate PHI leakage meeting HIPAA Safe Harbor and Expert Determination de-identification rule. • Preliminary evaluation demonstrating the accuracy of PHI detection and the efficiency of the system. 2. Background In this section, we provide background on HIPAA, explore Agentic AI systems in healthcare, and discuss the Access Controls. 2.1. Health Insurance Portability and Accountability Act (HIPAA) HIPAA, enacted in 1996, is a landmark federal law in the United States designed to protect PHI, ensure continuity of health insurance coverage, and modernize healthcare data exchange. The regulatory framework of HIPAA, codified in Title 45 of the Code of Federal Regulations (CFR), imposes stringent requirements on covered entities (e.g. healthcare providers, insurers) and their business associates to protect patient privacy and secure health data. Its provisions are particularly critical in an era of EHR and advanced technologies, such as Artificial Intelligence (AI) and LLM, which process vast amounts of sensitive health data. 2.1.1. Core Components of HIPAA HIPAA mandates comprise multiple regulatory rules, each governing distinct facets of PHI management and oversight. Privacy Rule (45 CFR § 160 and 164, Subpart E) : The privacy rule establishes standards for the use and disclosure of PHI, defined as individually identifiable health information transmitted or maintained in any form (45 CFR § 160.103). The key provisions include (i) Minimum Necessary Standard which means covered entities must limit the access, use, or disclosure of PHI to the minimum necessary to achieve the intended purpose (45 CFR § 164.502 (b)). (ii) Patient Rights - individuals retain the right to access (§ 164.524), request amendments (§ 164.526) and obtain an accounting of disclosures (§ 164.528) of their PHI. (iii) Authorization Requirements which entails written patient consent is mandatory for non-routine disclosures unrelated to treatment, payment or healthcare operations (45 CFR § 164.508). Security Rule (45 CFR § 164.302–318, Subpart C): The security rule establishes mandatory protections for electronic Protected Health Information (ePHI), requiring covered entities to implement a comprehensive framework across three core safeguard categories. Administrative Safeguards focus on organizational policies, including Risk Analysis Requirements (§ 164.308(a)), workforce security training (§ 164.308(a)(5)), and access control protocols (§

Read the complete text