What is Cowork? Use Cases Blog Services AI Readiness Quiz FAQ Book a Call

Compliance Guide · US Healthcare

HIPAA and AI: Can US Healthcare Providers Use Claude or ChatGPT in 2026?

A plain-English guide to Business Associate Agreements, de-identification, OCR enforcement, and the state laws that layer on top of HIPAA when generative AI enters the exam room.

Published November 13, 2026 · 9 min read

Short answer: yes, US healthcare providers can use Claude and ChatGPT under HIPAA — but only on BAA-covered plans for anything touching Protected Health Information, and only after your practice has a written data-classification policy, a tool allowlist, and trained staff behind it. Everything else in this guide is how to get there.

HIPAA Privacy Rule meets generative AI

The HIPAA Privacy Rule governs Protected Health Information (PHI): any individually identifiable health information a covered entity creates, receives, maintains, or transmits. That includes the obvious (name, date of birth, MRN, insurance ID, diagnosis, treatment notes) and the less obvious (ZIP code combined with age, appointment times, device identifiers, voice recordings).

When a clinician or admin pastes PHI into a chat window, HHS treats it as a disclosure. If the AI vendor is not a Business Associate under a signed agreement, that disclosure is impermissible — full stop. It does not matter that the data was typed rather than emailed, that the chat was private, or that the model "forgets" after the session. The disclosure happened when the characters left your network.

The practical result: every generative AI prompt your team writes is a mini risk assessment. Either the destination is covered by a BAA and the data can include PHI, or it is not and the data must be de-identified before it arrives.

Business Associate Agreements: who offers one in 2026

A Business Associate Agreement is the contract that lets a vendor touch PHI on your behalf. It imposes HIPAA Security Rule duties, breach notification timelines, and audit rights. Without one, the vendor is a stranger in HIPAA's eyes.

A BAA is required whenever the AI vendor will process PHI. If your workflow strictly uses de-identified data, a BAA is not strictly mandated — but most practices find it simpler to standardize on a BAA-covered tool than to police the line prompt by prompt.

De-identification: Safe Harbor vs Expert Determination

HIPAA recognizes two paths to lawfully de-identified data. Once data is de-identified correctly, HIPAA no longer applies to it — so it can safely flow to any AI tool.

Safe Harbor method

Remove all 18 categories of identifiers and confirm you have no actual knowledge the remaining information could re-identify someone. The 18 identifiers are:

Expert Determination method

A qualified statistician certifies that the risk of re-identification is "very small." Useful when you need to keep more granular data (e.g. full dates for a clinical research dataset) and are prepared to pay for the analysis.

Practical anonymization pattern for prompts: replace the patient's name with [Patient], DOB with [age X], address with region only, MRN with [Patient ID], and any dates with relative time ("2 weeks post-op"). Review the prompt out loud before sending. If a stranger could identify the person from what's left, keep redacting.

OCR enforcement posture in 2024–2026

The HHS Office for Civil Rights (OCR) has not yet published an AI-specific breach settlement, but its 2024 and 2025 guidance is unambiguous: generative AI vendors that touch PHI are business associates and the usual rules apply. OCR's 2024 bulletin on tracking technologies — which has driven seven-figure settlements against health systems that let third-party scripts receive patient data — applies the same analytical framework. A consumer AI tool receiving PHI is, legally, the same shape of problem as a web analytics pixel receiving PHI.

Expect the first AI-specific OCR settlements to follow the tracking tech pattern: multi-million-dollar resolutions with a vendor-management corrective action plan attached.

State laws that layer on top

If your patient base spans multiple states, the strictest overlay effectively sets your floor.

Compliance Framework

A 5-step compliance framework for US healthcare practices

Use this as the backbone of your AI governance policy. We teach practices exactly these steps in our implementation engagements.

1
Classify your data

Write a one-page policy that defines three tiers: PHI, de-identified health information, and non-PHI business content. Every AI use case is tagged with a tier before work begins.

2
Approve a BAA-covered platform

Pick one or two AI tools that will be your official PHI-cleared systems (Claude Enterprise, ChatGPT Enterprise, Copilot under M365, Gemini in Workspace). Execute the BAA before any PHI is entered.

3
Anonymization workflow for non-BAA tools

For general-purpose prompts (drafting marketing, SOPs, training content), document the de-identification pattern staff must apply. Supply redaction templates and spot-check them.

4
Train everyone who types

HIPAA training now has to cover AI specifically: the BAA vs non-BAA distinction, the 18 identifiers, and what a "prompt breach" looks like. Document attendance for OCR.

5
Audit log and incident process

Turn on your AI tool's enterprise audit logs. Add "AI prompt disclosure" to your incident response plan so a mispaste is triaged under HIPAA's 60-day breach clock, not discovered months later.

FAQ

HIPAA and AI: common questions

Can US healthcare providers use Claude or ChatGPT under HIPAA?

Yes, but only on the right plan. HIPAA requires a signed Business Associate Agreement with any vendor that creates, receives, maintains, or transmits PHI on your behalf. Anthropic offers a BAA on Claude for Work (Enterprise); OpenAI offers one on ChatGPT Enterprise, ChatGPT Team with a BAA, and the API. Consumer and free plans are not HIPAA-covered.

Is pasting a patient summary into ChatGPT a HIPAA breach?

If the plan is not covered by a BAA, pasting identifiable PHI into the tool is an impermissible disclosure under the Privacy Rule. It is treated the same as emailing that information to an uncovered third party and triggers breach notification obligations if risk cannot be ruled out through your documented risk assessment.

What are the 18 HIPAA identifiers I need to strip out?

The Safe Harbor method requires removing names; geographic subdivisions smaller than a state; all date elements (except year) directly related to an individual; phone and fax numbers; email addresses; SSNs; MRNs; health plan beneficiary numbers; account numbers; certificate/license numbers; vehicle IDs; device IDs; URLs; IP addresses; biometric IDs; full-face photos; and any other unique identifying number, characteristic, or code.

Has OCR fined anyone over AI tool misuse yet?

As of late 2026, OCR has not published an AI-specific settlement, but its 2024–2025 guidance treats generative AI vendors as business associates whenever they touch PHI. Recent settlements involving tracking technologies and unauthorized vendor disclosures apply the same logic — practices should assume OCR will treat an uncovered AI tool exactly as any other uncovered cloud vendor.

Do California, Texas, or New York add rules on top of HIPAA?

Yes. California's CMIA extends protection to any provider or contractor handling medical information, with broader private-right-of-action exposure. Texas HB 300 applies a wider definition of covered entity and requires biennial training. New York's SHIELD Act applies reasonable-safeguards and breach-notification duties to any business holding New Yorkers' private information.

What is the single most important control to put in place first?

Data classification combined with a tool allowlist. Decide which AI tools are approved for PHI (BAA-covered only) and which are only approved for de-identified or non-PHI work, then train every staff member on the difference. Nine of ten AI compliance incidents we see start with someone pasting PHI into the wrong tab.

Build the HIPAA-safe AI stack for your practice

We help US healthcare practices stand up compliant AI workflows — policy, platform selection, prompt libraries, and staff training. All in weeks, not quarters.

See Services Book a Discovery Call

Training from $149 per person · Team (up to 5) $349 · Implementation from $2,000