Is ChatGPT HIPAA compliant?

The short answer is no for the ChatGPT plans most practices and staff actually use, and yes for a handful of plans that come with a signed Business Associate Agreement (BAA). The difference matters because HIPAA doesn’t certify software. It requires that any vendor receiving protected health information (PHI) on your behalf signs a BAA, and that you use the product in a compliant way. This guide covers which OpenAI products qualify as of November 2026, what they exclude, and what you still have to do yourself.

Which ChatGPT plans are HIPAA eligible

The plan list comes from OpenAI’s HIPAA implementation guide (July 2026), which states that “ChatGPT services for consumers (e.g., Free, Plus, Pro) and ChatGPT Business are not Eligible Services.” Without a BAA, the same guide says, “you may not use the Services with Protected Health Information.”

Why “no training” isn’t the same as HIPAA compliant

This is the most common misunderstanding about ChatGPT and HIPAA. ChatGPT Business, Enterprise and the API don’t use your data to train models by default, and consumer users can switch off “Improve the model for everyone.” People take that to mean the data is safe for patient information.

HIPAA doesn’t ask whether a vendor trains on your data. It asks whether the vendor is a business associate with a signed agreement. Under 45 CFR 160.103, any company that “creates, receives, maintains, or transmits” PHI on your behalf is a business associate, and 45 CFR 164.502(e) requires a BAA with it. HHS has applied this even to cloud providers that only store encrypted data they can’t read. An AI provider reading your prompts is plainly a business associate.

So a Business plan with training turned off is still a vendor without a BAA. Putting PHI into it is a disclosure HIPAA doesn’t permit.

What’s still excluded on eligible plans

A BAA covers specific services and features, not everything in the product. On HIPAA-eligible workspaces, OpenAI preconfigures settings that limit features that don’t support compliance. Based on OpenAI’s guide and help center as of November 2026:

• Not covered: early-access and beta features, improved memory, event-triggered scheduled tasks, Sites, Codex Cloud, and internet access inside canvas and code blocks.
• Never put PHI in: user profile names and emails, workspace names and images, or support tickets and screenshots sent to OpenAI.
• Connectors and apps can send data to third parties that don’t offer HIPAA-covered services. Turn off the ones you don’t need.
• Sharing of conversations, projects and GPTs should only be on if every user who can see them is allowed to see that PHI.
• API: the BAA only applies when Modified Retention (for example, zero data retention) is active for each organization and project that handles PHI.

OpenAI’s help center lists covered features for each eligible product. Anything not on that list should be treated as not covered. We couldn’t confirm from OpenAI’s published pages whether voice mode and general file uploads are covered, so check the current list or ask your account team before using them with patient data.

How to get a BAA from OpenAI

ChatGPT Enterprise, Edu or ChatGPT for Healthcare. These are sales-managed. Contact OpenAI sales, and make sure the workspace is set up as a Regulated Workspace. OpenAI’s help center says it doesn’t offer a BAA for ChatGPT Business, so upgrading from Business means moving to Enterprise.

ChatGPT for Clinicians. Free for verified US physicians, nurse practitioners, physician assistants and pharmacists (verified by NPI number), with a BAA available for eligible accounts. It’s built for individual clinicians, not for a whole practice’s admin staff.

OpenAI API. Since October 2026, eligible organizations can accept OpenAI’s standard BAA in the API platform settings under the HIPAA compliance option. Accounts shown as “not eligible yet” need to go through OpenAI, and custom BAAs go to [email protected]. Accepting the BAA and enabling Modified Retention are separate steps; check that both are active for the project that will touch PHI.

What you still own after signing

The BAA moves some responsibility to OpenAI, not all of it. OpenAI’s own guide says that following it “does not ensure compliance.” Your side includes:

• A risk analysis that includes the AI tool (45 CFR 164.308).
• Access control: single sign-on, removing departed staff, and limiting who can use the PHI workspace.
• Minimum necessary: staff should enter only the information the task needs.
• Device security for any laptop or phone that opens the workspace.
• Retention and deletion. On ChatGPT Enterprise, OpenAI keeps customer content for the term of the agreement unless you delete it.
• Training staff on which workspace is approved for PHI, and which isn’t.
• Not using ChatGPT as a record system. OpenAI’s guide says it shouldn’t be used as a designated record set; the chart stays in your EHR.

If your staff are already using ChatGPT

In many practices, staff are already using personal or Business ChatGPT accounts for letters, summaries and emails. If that’s you, three steps:

1. Find out what’s in use. Ask, and check expense reports for ChatGPT subscriptions.
2. Treat past PHI entries as possible breaches. Under 45 CFR 164.402, an impermissible disclosure is presumed to be a breach unless a documented risk assessment shows a low probability of compromise. Record what was entered, delete the conversations, and run the assessment.
3. Give people an approved option. Bans without an alternative don’t hold. Either move to an eligible plan with a BAA, or give staff a tool that runs inside your own cloud account.

When a private setup makes more sense

For many small practices, ChatGPT Enterprise is more than they need, and Business isn’t allowed. Another route is running a model inside your own cloud account. Amazon Bedrock, for example, is on AWS’s HIPAA-eligible services list, is covered by the AWS BAA, and doesn’t share your data with the model providers. Your patient data stays in an account you control, and you can connect it to your EHR, intake forms and document storage.

That’s the setup we build: private AI in your own cloud, connected to the systems you already use. See AI consulting or start with the AI audit. For the bigger picture across ChatGPT, Claude, Gemini and Copilot, read our guide to HIPAA-compliant AI, and if you’re weighing Anthropic’s tools, see is Claude HIPAA compliant.

This guide is general information, not legal advice. Plan names and covered features change; confirm the current list with OpenAI before you rely on it.

Sources

• OpenAI HIPAA Implementation and Configuration Guide (July 2026): https://cdn.openai.com/osa/baa-hipaa-guide.pdf
• OpenAI help center, HIPAA-eligible products and functionality: https://help.openai.com/en/articles/20001069-hipaa-eligible-products-and-functionality
• OpenAI help center, getting a BAA with OpenAI: https://help.openai.com/en/articles/8660679-how-can-i-get-a-business-associate-agreement-baa-with-openai
• OpenAI enterprise privacy: https://openai.com/enterprise-privacy/
• ChatGPT for Clinicians launch coverage, Fierce Healthcare: https://www.fiercehealthcare.com/ai-and-machine-learning/openai-launches-chatgpt-clinicians-free-ai-tool-physicians-nps-and
• OpenAI for Healthcare: https://openai.com/index/openai-for-healthcare/
• 45 CFR 160.103 (business associate definition): https://www.law.cornell.edu/cfr/text/45/160.103
• 45 CFR 164.402 (breach definition and risk assessment): https://www.law.cornell.edu/cfr/text/45/164.402
• 45 CFR 164.514 (de-identification Safe Harbor): https://www.law.cornell.edu/cfr/text/45/164.514
• HHS guidance on HIPAA and cloud computing: https://www.hhs.gov/hipaa/for-professionals/special-topics/health-information-technology/cloud-computing/index.html
• AWS HIPAA eligible services reference: https://aws.amazon.com/compliance/hipaa-eligible-services-reference/
• Amazon Bedrock security and compliance: https://aws.amazon.com/bedrock/security-compliance/