Security & Compliance

Security for On-Premise Healthcare AI

HIPAA-friendly by architecture. On-premise by default. Auditable end to end. Written for your security team, your privacy officer, and your compliance committee.

How does Gofar Labs support a hospital’s HIPAA review?

Gofar Labs is designed with controls that support a hospital’s compliance process, but compliance is specific to each deployment and hospital review. The platform is described as HIPAA-friendly by architecture—not as a blanket certification or guarantee.

Gofar Labs on-premise architecture Inside the hospital network, clinical documents and existing systems feed the Gofar Labs server, which runs open-weight models on hospital GPUs. Staff review every draft in a review interface, and every request is written to an encrypted audit log. Outside the network, cloud AI services are not part of the data path; outbound traffic can be blocked, and updates arrive as signed offline packages. YOUR HOSPITAL NETWORK Clinical documents PDFs, faxes, scans, referral packets Existing systems EHR exports, scheduling, other MCP endpoints Gofar Labs server Open-weight models on your GPU · document processing · MCP gateway GPU Staff review Approve · edit · reject every AI draft before it is used Audit log Every request: actor, time, data scope · encrypted at rest · streams to your SIEM Cloud AI APIs & external model providers Not in the data path No outbound dependency Egress can be blocked at your firewall Updates Signed offline packages, on your schedule
The platform is designed to run with no outbound dependency: cloud AI services are not in the data path, and updates arrive as signed offline packages.

The Perimeter

No outbound dependency during normal operation.

The platform is designed to run without any outbound network dependencies. No telemetry, no cloud model calls, no shadow API endpoints. Every prompt, every completion, every tool invocation stays inside your building.

Your network team can put the Gofar Labs subnet behind a strict egress firewall. Nothing about the platform’s operation requires internet access. Software updates arrive through a signed, offline package process on a schedule you control.

Controls

Standard security primitives, from day one.

Encryption at rest

All persisted data on the Gofar Labs box is encrypted at rest using standard block-level encryption. Keys stay inside your infrastructure.

Role-based access control

Users and service accounts see only what their role permits. Access maps to your existing directory and IdP where possible.

Full audit logging

Every inference, every completion, every tool call is logged with actor, timestamp, and data scope. Streams into your SIEM.

Sensitive data domains

Department isolation for 42 CFR Part 2.

Behavioral-health data under 42 CFR Part 2 is held to a stricter standard than general HIPAA. Our multi-model architecture supports full department isolation: separate fine-tuned adapters so weights never mix information domains, separate audit trails, separate access rules.

A router model in front classifies each request and directs it to the correct isolated model. Information domains stay separate, all the way down to the adapter file.

Agent-to-agent traffic

The MCP gateway is a control point.

Hospitals are accumulating AI agents from multiple vendors. Without a control point, agent-to-agent traffic becomes an ungovernable back channel. The MCP gateway enforces which agent can call which tool, with what data scope, at what rate — and logs everything.

Your compliance team sees every interaction. Your security team can revoke any agent’s access with one policy change. The gateway becomes the auditable seam between all the AI vendors in your building.

A note on wording

We describe the Gofar Labs stack as “HIPAA-friendly by architecture.” We don’t call the platform “HIPAA compliant” — compliance is a property of a specific deployment inside a specific hospital, signed off after that hospital’s own audit. What we ship is a stack designed to make your compliance work possible: on-premise inference, no external calls, encryption, audit logging, role-based access, and department isolation. The last-mile compliance work is done together, with your team.

Common questions

Is Gofar Labs HIPAA compliant?

We describe the stack as HIPAA-friendly by architecture. HIPAA compliance is a property of a specific deployment inside a specific hospital, confirmed through that hospital’s own review. The platform provides the controls that review looks for: on-premise inference, encryption at rest, role-based access, and audit logging.

Can the platform run without internet access?

Yes. It is designed to run with no outbound network dependencies — no telemetry and no cloud model calls. Software updates arrive through a signed, offline package process on a schedule you control.

What does the audit log record?

Every inference, completion, and tool call, with actor, timestamp, and data scope. Logs can stream into your SIEM, and the MCP gateway records agent-to-agent calls the same way.

How is 42 CFR Part 2 behavioral-health data kept separate?

Through department isolation: separate fine-tuned adapters per information domain, with a router model directing each request to the correct isolated model, so weights never mix domains.

Can our security team review the code?

The platform code that runs inference, audit logging, and access control is meant to be auditable by any security team. See what we open-source and what stays private.

Planning a private LLM? Our private LLM deployment guide lists the governance questions to bring to any vendor, including us.

Comparing vendors? Read our buyer’s guide: On-premise vs cloud AI for hospitals — which vendors deploy behind your firewall.

Running several locations? Read: HIPAA firewall rules for multiple locations — and what AI changes.

Bring your security team to the call.

We prefer conversations where the CISO, privacy officer, and network team are in the room. It’s faster to answer real questions than to guess at objections.