NewSolo access is available. Invite codes are issued on request.Redeem invite →
Healthcare

Don't let a paste box create your next breach notification.

PHI-grade AI belongs inside your own perimeter. PrivateStack Enterprise deploys Llama, Mistral, DeepSeek, and other open-weight models into your own cloud environment: HIPAA-ready architecture, full audit trail, and PHI that never leaves your VPC.

Why it matters

Why healthcare teams choose PrivateStack

The risk

Paste a patient case into ChatGPT and you've made an unauthorized disclosure. No exceptions, no gray area.

With PrivateStack
PrivateStack gives every conversation a governed, fully logged home. For PHI workloads, an Enterprise deployment inside your own cloud environment keeps PHI under the access controls and safeguards you already have in place.
HIPAA-ready architecture, SOC 2 Type II in progress. We'll show you exactly where that stands and when it's expected to close.
The risk

Every SaaS AI vendor that touches PHI becomes another business associate to paper, audit, and worry about.

With PrivateStack
On an Enterprise deployment, PHI never leaves your VPC and never reaches PrivateStack systems, so your business-associate surface stays inside your own cloud. Bring your privacy officer to the security review. We'll walk the data-flow diagram together and work through BAA requirements where they apply.
On an Enterprise deployment there are no API calls to outside model providers. Nothing leaves your AWS account, so it runs inside the risk assessment cycle you already have, not a new one.
Your clinical documentation and internal knowledge base, searchable through AI inside your firewall, so staff get real answers without exporting PHI to get them.

At a glance

  • PrivateStack is a governed AI workspace for health systems, with HIPAA-ready architecture and configurable retention.
  • Models: open-weight families including Llama, Mistral, DeepSeek, GLM, Qwen, and Gemma.
  • Every request is logged in a complete, exportable audit trail: who accessed what, and when.
  • Hosted-tier inference is processed by a disclosed US-based subprocessor under contractual zero-retention terms: prompts are never stored and never used for training.
  • Enterprise BYOC deploys inside your own cloud account. On a BYOC deployment, PHI never leaves your VPC.
  • Pricing: Solo / Entrepreneur is $49/month for one seat, by invitation. Team is $99/user/month for up to 9 seats, sales-assisted. SOC 2 Type II certification in progress.

Plan the evaluation

Bring the workflow, the owner, and the evidence.

Synthetic example — not a customer story or measured result.

Draft a staff-facing administrative policy from fictional, PHI-free notes. Keep patient information out of the exercise and have a human review the draft.

Who decides
Privacy officer, security lead, and workflow owner
Data risk to resolve
A useful administrative exercise could expand into patient information without a fresh privacy and deployment review.
Deployment question
If the workflow later touches PHI, what deployment boundary, agreements, and access design would your privacy officer require before that change?
Evidence to request before approval
  • A written PHI-free scope for the exercise
  • A data-flow and subprocessor inventory for the proposed configuration
  • A privacy review of agreements, access, and retention before any PHI use

Use this checklist to prepare your conversation. It is not evidence that a deployment passed review. Confirm current controls against the Trust Center and the configuration being proposed.

Review the healthcare vendor evidence checklist

Guides for your AI evaluation

Review the questions for your team before choosing a workspace.

  • AI Vendor Checklist for HIPAA-Regulated Teams: 14 Checks Across the Full PHI Path

    Fourteen checks for evaluating any AI vendor that will touch protected health information, built around tracing the full PHI path — prompts and uploads, inference, storage, embeddings and derived data, audit logs, and deletion at termination. Business-associate status first (including no-view services), satellite agreements down the subprocessor chain, the embeddings question competing checklists skip, and the 60-day breach-notification clocks of 45 CFR 164.404 and 164.410 — every item cited to the instrument's own text.

  • Zero Data Retention vs No Training: What Each One Actually Covers

    Zero retention and no training are separate promises on separate axes. A map of the eight planes a single AI request can leave content on, who controls each clock, and the question to ask a vendor about each.

  • AI Vendor Security Questionnaire: 21 Questions, Scored (XLSX + Web Tool)

    A vendor security questionnaire built for AI: 21 questions across seven domains — data flow and subprocessors, retention and deletion, logging and export, isolation and deployment boundary, model and training use, incident response, access control. Per question: what a good answer looks like, required evidence, the red flag, a compensating control, and the contract clause. Score it in the browser or download the XLSX.

Ready to see PrivateStack in your environment?

Open models running in your own cloud, with the logs and controls to prove it. We'll walk your team through the exact architecture.