Is ChatGPT HIPAA Compliant? 2026 BAA Guide by Plan 

‍

OpenAI does not offer a business associate agreement (BAA) for ChatGPT Free, Plus, Pro or Business, so patient data should never go into those plans. It does offer a BAA for ChatGPT Enterprise with a Regulated Workspace, ChatGPT for Healthcare, ChatGPT for Clinicians and the API with Modified Retention.

So is ChatGPT HIPAA compliant? The better question is whether your specific setup is. HIPAA does not certify software. It governs how healthcare organizations and their vendors handle protected health information, and a signed BAA is only the starting point. That matters for clinics, billing and legal teams, health-tech vendors and any employee who pastes notes into a chat window. In this guide, we cover BAA status by plan, what a BAA leaves out and a checklist that a small team can follow. We are not lawyers, so treat this as general information, not legal advice.

Is ChatGPT HIPAA compliant? The short answer by plan

Whether ChatGPT is HIPAA compliant depends on the plan and on a signed BAA. OpenAI's HIPAA eligible products page lists the products it makes available with a BAA, and its HIPAA implementation guide, posted in July 2026, says consumer ChatGPT and ChatGPT Business are not eligible services.

We have limited the table to what OpenAI itself lists as eligible or not.

OpenAI's BAA help article explains the process. For the API, you email OpenAI with your company details and use case, and an enterprise agreement is not required. For ChatGPT Enterprise or Edu, you work with sales. OpenAI reviews each request case by case, so approval is not automatic.

ChatGPT HIPAA requirements: what HHS expects

HHS does not approve or certify AI tools. Its guidance on cloud computing says the Office for Civil Rights does not endorse, certify or recommend specific technology or products. In our reading, no vendor can hand you a badge that settles the question.

What HHS does say is practical. A cloud provider that creates, receives, maintains or transmits PHI for a covered entity is a business associate, and the covered entity needs a business associate agreement before using it. Using the service without one is a violation for the covered entity or business associate. Our overview of HIPAA and AI governance covers how these roles apply to generative AI.

The minimum necessary standard matters for prompts. PHI should not be used or disclosed when it is not necessary for the purpose, and organizations should make reasonable efforts to limit it to what the task needs. A summary of a clinical note rarely needs the patient's name, address and date of birth.

If your team supports US healthcare clients as a vendor or subcontractor, you may be a business associate yourself. HHS says a subcontractor that handles PHI for a business associate is also a business associate. Whether your team sits in India or the USA, confirm the exact obligations with your contract and your counsel.

What a BAA does not cover

A BAA allows PHI into an eligible product. It does not make every feature in that product safe. OpenAI's own guide says that without a BAA you may not use its services with PHI, and it also says that following its guidance does not ensure compliance with your HIPAA obligations. Customers must evaluate their own use.

We would check three gaps first:

• Features outside the BAA. OpenAI's list of functionality not covered includes improved memory, event-triggered scheduled tasks, internet access in Canvas and code blocks, and browser and network access in the cloud. PHI should not be entered when using them. Features that are not listed are not automatically covered either.
• Configuration. Covered features apply only in an eligible workspace and for roles that have them enabled. OpenAI points to access management such as SSO and to confirming that every invited user is authorized for PHI.
• Where PHI goes in the product. OpenAI advises against PHI in profile fields, workspace names and support requests, including screenshots and attachments.

A HIPAA compliant ChatGPT deployment is therefore a configured workspace, not a plan name. It also depends on what your people do inside it.

The bigger risk: PHI in tools nobody approved

People use the AI tool that is open in front of them. A Wolters Kluwer Health survey of 518 healthcare professionals, run by CITE Research in December 2025 and published in January 2026, found that 40% had encountered unauthorized AI tools at work and nearly 20% admitted using them. One in 10 had used one for a direct patient care use case. A BAA cannot cover a tool that never goes through your agreement. 

A nurse, a billing coordinator or a paralegal who is short on time may paste a note into a personal ChatGPT account on the same laptop as an approved workspace. A BAA covers the workspace it was signed for. It does not follow that person to another account.

Detecting this is harder than it sounds. PHI is not only a name. It includes dates of birth, record numbers, insurance details and free-text clinical notes where a diagnosis is described in plain sentences. Pattern matching finds a record-number format. It misses a paragraph that identifies a patient by context. Our guide to PHI vs PII explains what counts as each.

In our view, the practical control is to classify what a prompt means before it leaves the device, then sanitize or block sensitive health data. Staff keep using AI, and PHI stays out of tools it should not reach.

How to use HIPAA compliant AI tools: a 6-step checklist

Using HIPAA compliant AI tools is a process. We recommend that a small team run these steps, which do not need a dedicated compliance function.

1. Decide what data is involved. Our checklist starts with the data, not the tool. List the tasks where staff want AI help and mark which include PHI. This week: ask each team which AI tools they already use.
2. Choose an eligible product. Match the task to a product on OpenAI's eligible list, or to another vendor that will sign a BAA. This week: retire any use of Free, Plus, Pro or Business for PHI.
3. Get the BAA in place first. No PHI goes in until the agreement is executed. This week: start the request, because OpenAI reviews each one case by case.
4. Configure the workspace. Set up SSO and roles, switch off features outside the BAA and keep PHI out of profile fields and support tickets. This week: switch off every feature on OpenAI's not-covered list for any workspace that handles PHI.
5. Limit what goes in. Apply the minimum necessary standard to prompts and uploads, and add a control that screens inputs. This week: write a one-page rule on what staff may paste.
6. Train, log and review. Staff need to know which account to use for what. Keep audit logs and review them on a schedule. This week: add a monthly review to the calendar.

How Wald AI keeps PHI out of unapproved AI tools

Wald AI protects sensitive business data across AI interactions, and PHI is one of the data types that matters most for healthcare teams. We built it as AI DLP: data loss prevention designed for how people use AI.

Instead of relying only on predefined patterns, small language models deployed locally on the device classify the meaning and intent of a prompt or document. When they detect sensitive health data, guardrails sanitize it before it reaches an AI system, and audit logs record the interaction. That lets staff keep using AI tools while PHI stays protected.

Here is how that looked in practice. A U.S. personal injury law firm that reviews thousands of pages of medical records each month used Wald AI to screen every prompt and upload. The firm's Wald AI deployment sanitized PHI, such as patient identifiers and record numbers, before requests reached the model. Its security and compliance teams gained visibility into users, models and applied policies, with audit logs for review.

Wald AI supports your own governance with controls and audit logs. It does not replace a BAA, a risk analysis or legal advice. To see how it works on the device, see Wald AI in action.

Is ChatGPT HIPAA compliant? The bottom line

Only in the right product, under a signed BAA, with the right configuration and with careful use. Free, Plus, Pro and Business plans are off the table for PHI. The eligible products still have features outside the BAA.

Our advice is to get the agreement in place, limit what goes in and control the prompt itself. That is what a HIPAA compliant ChatGPT setup looks like in practice. Do that, and AI becomes something a small healthcare team can adopt with confidence. We would start with step 2 this week. If you want to see how that works for your team, book a demo or talk to our team.

‍