Privacy Policy — HCC Coding Assistant

Effective date: 2026-11-06 · Penguin AI · Contact: [email protected]

What this service does

The HCC Coding Assistant analyzes de-identified medical chart documents and clinical notes to suggest ICD-10 diagnosis codes with CMS-HCC risk adjustment mappings, E/M visit levels, and clinical summaries. It is a coding assistant: every suggestion requires review by a qualified professional before any billing use.

The data rule this service is built around

This service accepts de-identified and redacted documents only. It must not be used with documents containing patient identifiers. This is enforced in software, not just policy: every uploaded document is screened by a three-layer detection system (pattern matching, a clinical PHI entity detector, and an AI review) against the HIPAA Safe Harbor identifier categories — names, dates of birth, record and member numbers, contact details, addresses, and the rest. Documents in which identifiers are detected are refused without being processed further, and the refusal names only the identifier types found, never the values.

What we receive, and what happens to it

• Documents you submit are read transiently to screen and, if they pass, to analyze them. Document content is processed in memory; the original file is not stored by this service after processing completes.
• Analysis results (suggested codes, evidence citations, summaries) and operational job records are retained for up to 7 days to support retrieval and retry, then deleted automatically.
• Reviewer decisions (a coder accepting or rejecting a suggested code) are retained longer as an audit record, attributed to the authenticated account that made them.
• Account data: you sign in with your Gwen account (Penguin AI's platform, gwen.penguinai.co). This service receives your account identifier and email from that sign-in. We never see your password.
• Usage and billing data: for each billable request we record the tool used, AI token counts and the computed cost, and send them with your email to your Gwen account so the usage is charged to your Gwen credits. Before processing we check your credit balance with Gwen. No document content or analysis results are sent to Gwen.

What we emit

Results are returned in de-identified form consistent with HIPAA Safe Harbor §164.514(b)(2): no names or identifiers, no demographic details, and no dates finer than the year — encounters are labelled by year and sequence (e.g. "2024, encounter 1"). Quoted evidence from the document is swept for dates before being returned.

Where processing happens (subprocessors)

• Amazon Web Services — all document analysis (including AI model inference via Amazon Bedrock and PHI screening via Amazon Comprehend Medical) runs inside our own AWS environment. Content submitted for analysis is not retained or used for model training by these services.
• OpenAI / ChatGPT — when you use this service through ChatGPT, your conversation and any files you attach are handled by OpenAI under OpenAI's own terms and privacy policy before and independently of this service. Files you attach are stored by OpenAI and fetched by this service for screening.
• Gwen (Penguin AI) — sign-in, account identity and credit billing (email, tool name, token counts and cost per request). Gwen is operated by Penguin AI. No document content or analysis results are sent to it.

What we do not do

• We do not sell or share your data, or use it for advertising.
• We do not use your documents to train models.
• We do not knowingly process protected health information; documents found to contain identifiers are refused.
• Suggestions are not medical or billing advice; a qualified coder must review them before use.

Security

All connections use TLS. API access requires OAuth 2.1 sign-in; every reviewer action is attributed to the verified account that performed it. Cloud resources are authenticated by short-lived workload identity, not stored credentials.

Your choices and contact

To ask questions, request deletion of retained records associated with your account, or report a concern, contact [email protected]. We answer within 5 business days. If we change this policy, the effective date above changes with it.