NewMissing Records Detection: flags every visit, provider, and date missing from the file. See how →
Home›Legal Center›Subprocessor List
Legal · Document 05

Subprocessor List

Every third party we engage to process Customer Content, what it does, where, and under what safeguards. Short list, on purpose: 4 vendors can touch your records, every one under a BAA with zero-retention, no-training terms. Your files are stored on AWS in US data centers. Every AI model request goes through one gateway, OpenRouter. Changes are announced 30 days in advance, and business-operations vendors that never touch Customer Content are listed separately.

Last updated: November 3, 2026 Change notice: 30 days in advance

A. Subprocessors of Customer Content

These entities can process the records, imaging, and outputs you upload, solely to provide the Services. Each is bound by a subcontractor BAA and terms no less protective than the DPA, including the no-training flow-down.

Entity
Service
Location
Safeguards
Amazon Web Services, Inc.
Cloud infrastructure: compute, storage, database, and monitoring for the platform (Self-Service plan)
United States
us-east-1 ·
us-west-2 (DR)
BAA executed; HIPAA-compliant data storage in US data centers; isolated VPCs; AES-256 at rest, TLS 1.2+ in transit
Anthropic, PBC
Commercial backup model, reached only through OpenRouter: clinical-event extraction, summarization, record Q&A, citation generation
United States
BAA executed; reached only through OpenRouter, not called directly; zero-retention inference endpoints; contractual no-training on Customer Content
OpenAI, LLC
Commercial backup model, reached only through OpenRouter: vision OCR for scans and handwriting, structured extraction
United States
BAA executed; reached only through OpenRouter, not called directly; zero-retention inference endpoints; contractual no-training on Customer Content
OpenRouter, Inc.
AI inference routing: the single gateway for every model request. Most requests run on open-source models through OpenRouter. The commercial models listed above are a backup for steps where accuracy matters most
United States
BAA executed; zero-retention routing (no prompt or output logging); contractual no-training; routes only to models and endpoints approved under this list

Plan scope. This table describes Self-Service and free first case processing. Under Enterprise On-Prem, no subprocessor receives Customer Content unless you explicitly enable it. Details: DPA Annex 3.

B. Business-operations processors no access to Customer Content

These vendors support running the business: billing, email, site analytics. They never receive medical records, imaging, or any Customer Content; they process only the visitor, account, and billing data described in the Privacy Notice.

Entity
Service
Location
Data involved
Stripe, Inc.
Payment processing for per-page billing and Order Form invoicing
United States
Billing contact, payment method, transaction amounts. Card data goes to Stripe directly, never to us
Google LLC (Workspace)
Business email, documents, and internal collaboration
United States
Correspondence you send us; support and sales communications
Google LLC (Analytics 4)
Website analytics on medrecords.ai: loads only after cookie consent
United States
Pseudonymous usage data per the Cookie Policy; IP anonymization enabled

What every Customer-Content subprocessor signs

Subcontractor BAA under 45 CFR 164.502(e), no weaker than ours with you
Zero-retention inference: no storage of prompts or outputs beyond the request
No training on Customer Content, including de-identified derivatives: the DPA §4.4 flow-down
US-only processing of Customer Content, breach notice to us within 24 hours, audit cooperation

We remain fully liable to you for every subprocessor's performance, as if it were our own (DPA §7.3).

Change notifications & objection rights

We add or replace a subprocessor of Customer Content only after 30 days' advance notice, by email to your account contacts. You may object on reasonable data-protection grounds within 14 days to privacy@medrecords.ai; if we can't resolve the objection, you may terminate the affected Services with a pro-rata refund of prepaid, unused fees (DPA §7.2). Emergency replacements for security reasons are notified as they happen, with the same objection rights.

Changelog

Date
Change
Nov 3, 2026
Clarified the model pipeline (v1.2): all AI model requests go through OpenRouter, most to open-source models, with Anthropic and OpenAI as a backup reached only through OpenRouter. Customer data is stored on AWS in US data centers. No subprocessor was added.
Jul 7, 2026
Added OpenRouter, Inc. (AI inference routing) to the Customer Content table (v1.1).
Jul 7, 2026
Initial publication (v1.0): AWS, Anthropic, OpenAI (Customer Content); Stripe, Google Workspace, Google Analytics (business operations).