Is Gemini HIPAA Compliant? 2026 Guide (BAA, Workspace, Vertex AI, and the Gaps)
Is Google Gemini HIPAA compliant? The 2026 answer per product surface — Workspace, Vertex AI, AI Studio, consumer — plus the gap a signed BAA does not close.
No — Gemini is not HIPAA compliant by default on any plan. HIPAA compliance for Gemini requires three things, in this order:
The rest of this guide walks through each Gemini surface and the controls a healthcare organization needs.

Google's HIPAA Implementation Guide for Google Workspace and Cloud Identity and the Google Cloud HIPAA Implementation Guide define the covered surface. A simplified summary for the AI-relevant products:
The single most common mistake: assuming a personal Gemini account ("free Gemini" or the Gemini app a clinician downloaded) is somehow covered because the same vendor offers a HIPAA-eligible product elsewhere. It is not. The BAA covers the surface, not the brand.
Source link: Google Workspace HIPAA Implementation Guide — verify against Google's current docs before signing.
Yes, on a properly configured covered SKU, with the BAA executed. The Workspace BAA covers Google Workspace Core Services as defined in Google's HIPAA Implementation Guide. Gemini for Workspace (the AI assistant inside Gmail, Docs, Sheets, Slides, Meet, Drive) is covered when the customer is on a SKU listed in that guide and has executed the BAA through the Admin console.
The healthcare org responsibilities are non-trivial:
That gives Google's BAA-covered processing. It does not give you visibility into what PHI users are actually pasting into a Gemini chat, attaching to a Gmail thread that triggers a Gemini summary, or storing in a Drive doc that Gemini will later retrieve as context.
Yes, under the Google Cloud BAA, when the project is configured for HIPAA. Vertex AI is included in the Google Cloud HIPAA-eligible services list. For healthcare ML/LLM workloads that need to call Gemini models directly, Vertex AI is the supported path.
Requirements include:
Vertex AI Gemini is the right path for healthcare engineering teams building applications that call Gemini directly. It is not the same as "all our employees can use Gemini for healthcare workflows" — that requires the Workspace path or a controlled application.
Only when accessed through a paid GCP project with the Cloud BAA in place. The free-tier Gemini API (via Google AI Studio API keys) is not under any BAA. If your application uses Gemini API and could touch PHI, it must run through Vertex AI on a BAA-covered Cloud project.
No. Google AI Studio (aistudio.google.com) is a free developer playground. It is not covered by any Google BAA and is not intended for production use, let alone PHI. Engineers prototyping a healthcare app should move off AI Studio to Vertex AI before connecting any real data.
No. Consumer Gemini (gemini.google.com under a personal Google account, the Gemini mobile app on a personal phone) is outside any BAA. A clinician who pastes a patient note into the Gemini app on their personal phone has transmitted PHI to a non-BAA-covered service.
This is the most-overlooked Gemini exposure in healthcare orgs. The org has Workspace under BAA — but employees use personal Gemini on the side. The data flow Google sees is from a personal account, not the covered Workspace tenant.
A BAA is a contractual document. It commits Google to handle PHI under HIPAA's Privacy, Security, and Breach Notification rules. It does not:
A BAA covers Google's processing of the data you sent. It does not minimize, redact, or restrict the data you send. That responsibility is on the covered entity. Most healthcare organizations only realize this gap after a near-miss.

Strac is the data-layer control that closes the gap a BAA doesn't cover. Strac sits between users and Gemini surfaces, between AI agents and Google Workspace data, and between endpoints and the model.
Browser extension for Gemini in the browser — Strac's GenAI browser DLP inspects copy/paste, typed text, and file uploads to gemini.google.com, the Gemini app in Workspace, and any other browser-based Gemini surface. Modes: alert, warn, block. PHI never crosses into the model context. Even on consumer Gemini, the browser extension enforces the same controls.
MCP DLP for Gemini-driven AI agents over Google Workspace data — Most modern AI workflows involve a Gemini-powered agent reading from Drive, Gmail, Calendar, Docs, and Sheets. Strac's Google Workspace MCP DLP intercepts those tool calls and redacts PHI inline:
Endpoint DLP for Gemini desktop and CLI tools — Strac's endpoint DLP on Mac and Windows catches file-level exfil that browser controls miss — for example, a Gemini code assistant integrated into VSCode pulling in source files with embedded credentials.
Cloud DSPM for Vertex AI applications — For applications calling Gemini through Vertex AI, Strac's cloud DSPM keeps the upstream data stores classified, so engineering teams know which datasets are HIPAA-sensitive before piping them into a Gemini prompt template.
Compliance evidence pre-built — HIPAA mapping, audit log export, and per-event remediation evidence ready for your auditor without a custom integration.
Setup is agentless and under 10 minutes per workspace.
Phase 1 — Inventory and BAA
Phase 2 — Data layer controls
Phase 3 — Governance and audit
No. Consumer Gemini is outside any Google BAA. Healthcare data sent to consumer Gemini is sent to a non-covered service.
Yes, on covered Workspace SKUs with the Workspace BAA executed. Verify the SKU against Google's current HIPAA Implementation Guide before signing.
Only the paid Gemini API accessed via Vertex AI on a Google Cloud project covered by the Cloud BAA. The free Gemini API via AI Studio API keys is not covered.
No. AI Studio is a free developer playground outside Google's HIPAA program. Move to Vertex AI before connecting any real healthcare data.
NotebookLM's coverage has shifted by plan tier. Check Google's current HIPAA Implementation Guide before using NotebookLM with PHI.
No. Google's BAAs require a paid Workspace plan with the BAA executed by an authorized admin, or a paid Cloud project. Free consumer accounts are outside the program.
The biggest practical difference is BAA coverage on the consumer-grade AI assistant. Google offers BAAs for Gemini inside Workspace and Vertex AI. Anthropic does not currently offer a BAA for Claude consumer or Claude Cowork plans — see Is Claude HIPAA compliant?. Both vendors leave the same data-layer gap that a BAA alone does not close.
Strac's browser DLP, endpoint DLP, and Google Workspace MCP DLP redact PHI before it reaches Gemini, regardless of which Gemini surface a user is on. A signed Google BAA covers Google's processing of the data you send. Strac controls what data gets sent in the first place. The combination — Google BAA + Strac data-layer protection — is what most healthcare auditors want to see.
Google's HIPAA Implementation Guide tells you how to configure covered Google products under the BAA. It is necessary. It is not sufficient. The Guide does not enforce what your users actually paste into Gemini, what AI agents retrieve from Drive, or what file content reaches Gemini via the browser. That layer is where Strac sits.
Yes. Strac's browser extension supports a block mode that prevents copy/paste and uploads into Gemini until you have the controls in place. Many healthcare orgs use a phased rollout: block in week one, warn-and-allow in week two after browser DLP is tuned, full rollout in week three.
Gemini can be deployed in HIPAA workflows — but only on covered Google surfaces with a signed BAA, and only with a data-layer control that minimizes PHI exposure on every surface in scope. Consumer Gemini and AI Studio are out. Workspace and Vertex AI are in, with the right SKUs and configuration. The remaining gap — what users paste, what agents retrieve, what files reach the model — is what Strac is built to close.
See how Strac protects Gemini in your environment — book a demo →
.avif)
.avif)
.avif)
.avif)
.avif)


.gif)

