AI in healthcare: Why Data Access Can't scale
AI agents turn healthcare data access from months-long projects into thousands of prompts. Why committee review can't scale to AI, and what replaces it.
subsalt.aiCrawled Nov 2, 2026
AI agents turn healthcare data access from months-long projects into thousands of prompts. Why committee review can't scale to AI, and what replaces it.
Subsalt provides health systems with controlled access to de-identified electronic health record data for AI, research, and collaborations, keeping protected health information in their systems.
The page explains how de-identification enables scalable access to health data for AI while adhering to HIPAA’s purpose-based governance.
Synthetic data for exploration. A secure runtime for production. HIPAA-compliant access to clinical data in minutes, not months.
Watch Governance at AI Speed: Health Data Access for Enterprise AI, with CHAI, Northwestern, and Subsalt.
Company news, product announcements, and press releases from Subsalt. Media enquiries: media@getsubsalt.com.
Subsalt Secure Runtime lets health systems authorize analysis involving PHI without exposing patient-level data. Datavant has certified Subsalt's de-identification across both halves of the platform under HIPAA's Expert Determination standard.
Senior Data / ML Engineer role at Subsalt. Co-own the privacy methodology that decides what data is safe to release, and build the pipeline that produces it, at a post-revenue data infrastructure startup.
The page describes a data infrastructure that balances speed and compliance for data and compliance teams.
The page offers a consultation on privacy-preserving strategies for health systems’ AI, research, or data-access projects.
Subsalt automates HIPAA-compliant de-identification so research and data teams access clinical data in minutes, not months.
Subsalt's privacy policy.
Subsalt's terms of service.
Subsalt introduces a **Secure Runtime** environment enabling health systems to authorize data analysis involving **protected health information (PHI)** without revealing individual patient records. Datavant assessed the platform’s privacy safeguards under **HIPAA’s Expert Determination standard**, validating compliance with strict data protection requirements. The solution aims to expand secure health data analytics while maintaining patient confidentiality.
Subsalt deploys a HIPAA-compliant synthetic data sandbox on Microsoft Azure and Fabric to enable secure medical research access. The platform reduces UT Southwestern’s clinical data retrieval from months to days, accelerating research workflows by 50 times. Researchers and students now access anonymized datasets without compromising patient privacy.
Subsalt transforms healthcare data compliance challenges into an accelerator for research. The company helps institutions navigate regulatory hurdles so they can focus on leveraging existing datasets for meaningful insights. By streamlining data access, Subsalt enables faster, more efficient use of patient and operational data in clinical and operational research.
Intel Ignite’s accelerator program has chosen **30 deep tech startups** for its **Spring 2023 cohorts** across the **US, Europe, and Israel**. The initiative, launched in 2019, provides early-stage companies with Intel’s resources to refine products, optimize valuations, and foster strategic growth. Past participants have collectively secured **over $1.5 billion** in funding and milestones through the program’s support.
Subsalt develops a hybrid approach to anonymize health data for research by combining techniques like data masking, aggregation, synthetic data, and differential privacy. The company addresses challenges in balancing data utility with protection of Protected Health Information (PHI). This blended solution aims to enable advanced analytics while mitigating risks associated with traditional anonymization methods.

Subsalt provides de-identification infrastructure for health systems under HIPAA Expert Determination, combining synthetic data for exploration with a publishable runtime.