No walled gardens. Use ChatGPT, Claude, Gemini - any model you want.
95% F1 accuracy removing PHI. 100% HIPAA compliant.
The first PHI scrubber designed specifically for AI interactions
MedScrub automatically detects and removes PHI from healthcare data before sending it to AI models like GPT-4, Claude, or Gemini. Then it restores the context in AI responses - giving you safe, compliant AI interactions.
Healthcare teams want to use AI for clinical insights, patient communication, and research. But HIPAA compliance stops them cold. MedScrub bridges this gap - enabling safe AI use without sacrificing patient privacy.
Unlike cloud-based solutions, MedScrub runs as a Docker container on your infrastructure. PHI never leaves your environment. You get AI capabilities without vendor liability or data exposure.
Created by developers building CLaiR (AI assistant for patient medical records) who faced the same HIPAA barriers. MedScrub emerged from real-world need, not theoretical compliance requirements.
Three simple steps to HIPAA-compliant AI
Type or paste medical text with PHI into our chat interface or send via API
MedScrub removes PHI, proxies to AI models (GPT-4, Claude, etc.), then restores context
Receive AI insights with PHI automatically restored, fully HIPAA compliant
Choose the deployment that fits your workflow
For healthcare teams
Get early access to MedScrub for your infrastructure
docker pull medscrub/api:latestexport MEDSCRUB_API_KEY=msk_live_your_api_key_heredocker run -d -p 8080:8080 -e MEDSCRUB_API_KEY medscrub/api:latestcurl -X POST http://localhost:8080/api/deidentify \
-H "X-API-Key: $MEDSCRUB_API_KEY" \
-H "Content-Type: application/json" \
-d '{"text": "Patient John Doe (SSN: 123-45-6789) visited on 1/15/2024"}'Powered by advanced AI with 96% F1 accuracy
Ensemble model combines pattern matching, NER, and contextual analysis for comprehensive coverage.
Detects complex international names, compound surnames, and cultural naming patterns.
Preserves medical terminology, drug names, and clinical values while removing PHI.
Sub-8ms response times with intelligent caching. Process thousands of requests per second.
Enable AI safely across your entire organization
Let AI help with clinical notes, discharge summaries, and care plans without HIPAA risk.
Analyze patient data patterns and literature reviews while protecting patient privacy.
Automate insurance letters, referrals, and prior authorizations with AI assistance.
Join healthcare organizations already using MedScrub to safely leverage AI
MedScrub combines multiple algorithmic approaches and compliance frameworks to achieve industry-leading accuracy in PHI detection
Comprehensive detection of all 18 HIPAA Safe Harbor identifiers ensures regulatory compliance
Fast, deterministic pattern matching with high-precision validation for structured identifiers
spaCy, BioBERT, ClinicalBERT for context-aware named entity recognition and medical understanding
Enhanced patterns inspired by UCSF's Philter system for comprehensive medical entity detection
Bloom filters and MinHash for fuzzy matching, probabilistic validation, and edge case handling
Multiple detection methods work together to achieve 96% F1 accuracy with minimal false positives