robotnurse.io

./medical-ai — private AI for small medical & dental practices

HIPAA-safe AI that never sees the cloud.

The wall a small practice hits before any technical question is the business associate agreement: the AI tools people reach for first won't sign one, so they can't touch patient data at all. Run the model on hardware the practice owns and the question collapses — no third party holds your patients' information. Built for practices in New York City and nearby New Jersey.

✎

RobotNurse sets up private, offline AI on hardware your practice owns. This page is about HIPAA and AI in a medical or dental office — it is not healthcare, nursing, or hospital robotics, and nothing here is legal or compliance advice. Confirm specifics with a healthcare attorney or qualified HIPAA consultant.

[ the gate that decides it first: the BAA ]

  1. 01

    No BAA, no PHI — full stop

    Under 45 CFR 164.502 and 164.504, a covered entity may hand PHI to another party only when that party is a business associate under a written BAA. The first question about any AI tool is therefore not “is it good” — it is “will they sign a BAA, and is the product even eligible for one.” If either half is no, the tool is off the table for anything containing PHI, regardless of capability.

  2. 02

    The tiers that trip people up

    Per OpenAI's own eligibility matrix, ChatGPT Free, Go, Plus, Pro, Team, and Business are all not BAA-eligible. Only sales-managed Enterprise and Edu, and the API under an executed BAA, can touch PHI. “Business” sounds like the compliance tier — it is not.

  3. 03

    “HIPAA compliant” is not a certification

    HHS's Office for Civil Rights runs no certification program for software. HIPAA is three rules — Privacy, Security, and Breach Notification — that impose obligations on you, the covered entity. A tool can't be compliant on its own; the question is whether the arrangement lets you meet your obligations.

[ what running it on your own hardware changes ]

the third party disappears when the model runs on hardware the practice owns and PHI never transits a public network, there is no third party receiving it. No third party means no business associate relationship, which means no BAA is required for that component. That is a real, categorical difference — and the whole reason on-prem AI is interesting to a small practice.
your obligations do not disappear the practice is still a covered entity and the Security Rule (45 CFR Part 164, Subpart C) still applies to the local system: documented risk analysis, access controls, audit logging, and encryption at rest.
the honest trade you exchange a vendor-trust problem for an operations problem. Patching, access control, backups, and physical security become yours — and sizing the box is its own decision, which the RAM calculator handles.

[ the compliance checklist ]

Physical & access controls

The machine lives in a locked space; only authorized staff have accounts, with per-user credentials and no shared logins.

Audit logging

Who ran what query, when. If you ever need to answer an OCR inquiry, this is the first thing they ask for.

Encryption at rest

Full-disk encryption (FileVault on macOS, LUKS on Linux), so a stolen or decommissioned drive isn't a breach.

Documented risk analysis

The Security Rule's foundational requirement, whether your AI is on-prem or in the cloud.

BAAs where they still apply

Your managed IT provider, your EHR, your billing service — every one of those still needs its own BAA. Local AI fixes the AI component, not the rest of the compliance program.

De-identification is real but narrow

Under the 45 CFR 164.514 safe harbor, stripping 18 specific identifiers makes data no longer PHI. In practice this almost never works for live notes, which are loaded with identifiers — it is a documented process, not something that happens by accident.

Human in the loop

A local model is weaker than the frontier on some clinical tasks. Match the task to the model (the model-fit tool helps) and keep a human on anything clinical.

[ the three-question test ]

  1. 01

    Does patient information leave the practice?

    If yes, is there a BAA with that vendor? If no BAA, the tool cannot touch PHI — capability is irrelevant.

  2. 02

    Who verifies the output?

    Name the human and the step. If the answer is “nobody,” the use is not compliant at any price.

  3. 03

    Who is accountable when it's wrong?

    If the answer is “the vendor,” stop. It is the covered entity, always.

A local model makes question 1 easy and does nothing for questions 2 and 3 — a fair summary of where on-prem AI actually helps a small practice.

HIPAA-SAFE AI FOR YOUR PRACTICENYC + NJ

If the answer to question 1 is “yes, and that worries us,” the next step is a conversation, not a purchase. Tell me about your practice and I'll map the Security Rule onto your actual systems.

book a free scoping call   start with the $899 readiness audit →   see the services →   how local AI setup works →