Which AI tools will sign a BAA, and on which plan
Which AI tools can a clinic use with patient data? A plain guide to which ChatGPT, Claude, Copilot and Gemini plans come with a HIPAA BAA, and what to check.
Shane Donoher · Published November 2, 2026
The short answer
A clinic can only put patient information into an AI tool when the vendor has signed a Business Associate Agreement (BAA) with you, and only on the specific plan and features that BAA covers. The big four, OpenAI, Anthropic, Microsoft and Google, all offer a BAA for some of their AI products. None of them cover the free or personal versions your staff may already be using.
So the real question isn't "which AI is HIPAA compliant". It's "which plan are we on, has someone actually signed the BAA, and is it switched on the way the vendor says it has to be".
What a BAA is, in one paragraph
Under HIPAA, any outside company that creates, receives, keeps or sends protected health information (PHI) for you is a business associate. Before you share PHI with them, you need a signed agreement setting out how they'll protect it. An AI tool that reads a patient note, a referral letter or a schedule with names on it is handling PHI. No BAA, no patient data. It's that simple.
Where the main vendors stand
This is based on each vendor's own public pages, checked on 2 November 2026. These terms change often, so check the current version before you rely on it.
OpenAI (ChatGPT). OpenAI offers a BAA for ChatGPT Enterprise and Edu with what it calls a Regulated Workspace, ChatGPT for Healthcare, ChatGPT for Clinicians, and its API with Modified Retention. It says plainly that ChatGPT Business, the plan many small teams buy, is not covered, and neither are Free, Plus or Pro. For Enterprise you go through their sales team.
Anthropic (Claude). Anthropic offers a BAA for Claude Enterprise plans that have been made HIPAA-ready and for its API. On Enterprise, the organisation's Primary Owner has to switch on the HIPAA setting and accept the BAA; a standard Enterprise plan isn't covered until they do. Some features, including several marked beta, are excluded.
Microsoft (Copilot). Microsoft's BAA comes by default through its Data Protection Addendum for customers who are covered entities or business associates, and it lists the services in scope. Microsoft 365 Copilot and Copilot Chat used with work accounts are among them, but parts of the experience, such as web search queries, are treated differently. Personal Microsoft accounts are a different matter entirely.
Google (Gemini). Google Workspace customers accept a BAA through the admin console. As of 31 August 2026, Google lists Gemini in Workspace and the Gemini app (excluding Gemini in Chrome) as covered functionality. Third-party add-ons are not covered, and a personal Gmail account is not covered at all.
The plan trap
For a lot of clinics, the gap isn't "no BAA". It's "wrong plan".
A practice manager signs up for a team plan with a company card. It looks like a business product. It has an admin panel. But the BAA only exists on a higher tier, or only after someone ticks a box in settings, or only for certain features. Meanwhile, half the front desk is using the free version on their phones because it's quicker.
That's why the first job is finding out what's actually in use, not what's on the invoice. Our post on finding where your team's AI data is going covers how.
A signed BAA isn't the finish line
Every one of these vendors says the same thing in its own words: a BAA supports your compliance, it doesn't give it to you. Once it's signed, you still need to:
- set the workspace up the way the vendor's HIPAA guide says, which often means turning some features off
- keep PHI out of the places the vendor excludes, such as profile names, support tickets, beta features and third-party connectors
- decide who on the team is allowed to use AI with patient information, and for what
- make sure a person checks anything AI writes before it goes into a chart or out to a patient
- keep a record of all this, so you can show it if anyone asks
Five steps for a clinic this month
- List what's in use. Ask staff which AI tools they've tried, on which devices and which accounts. Make it safe to answer honestly.
- Check the plan, not the brand. For each tool, find the exact plan name and check it against the vendor's current HIPAA page.
- Get the BAA signed, or stop. If a tool has no BAA on your plan, PHI stays out of it until that changes.
- Switch on the settings. Follow the vendor's HIPAA setup guide and turn off features it excludes.
- Write one page of rules. Which tool, for what, who checks the output, and what never goes in. Then show the team.
If you'd like someone to do steps one to four with you, that's what our AI Safety Check is for. We find out how your team is already using AI, lock it down with a Safe AI Setup, and keep it safe every month with Managed Safe AI, because vendor terms like these keep moving.
FAQs
Is ChatGPT HIPAA compliant?
No AI tool is "HIPAA compliant" on its own. OpenAI will sign a BAA for certain products, including ChatGPT Enterprise with a Regulated Workspace, ChatGPT for Healthcare, ChatGPT for Clinicians and its API with Modified Retention. It does not offer one for ChatGPT Business, Free, Plus or Pro.
Can we use Microsoft Copilot with patient data?
Microsoft lists Microsoft 365 Copilot and Copilot Chat for work accounts within the scope of its BAA, which comes through its Data Protection Addendum. Check the current list of in-scope services and how web search is handled, and keep personal Microsoft accounts away from patient data.
Does Gemini in Google Workspace come with a BAA?
Google lists Gemini in Workspace and the Gemini app, excluding Gemini in Chrome, as covered under its Workspace BAA, which an admin has to accept. Personal Gmail accounts and third-party add-ons are not covered.
If we have a BAA, can staff paste anything in?
No. Vendors exclude certain features and fields, and HIPAA's minimum necessary rule still applies. Share only what the task needs, use only covered features, and have a person check the output.
What if someone already put patient information into a free AI tool?
Don't panic and don't ignore it. Write down what was shared, in which tool and when, and talk to your privacy officer or legal adviser about whether it needs to be treated as a breach. Then fix the gap that let it happen.
*This post is general information, not legal advice. Vendor terms change often, so check each vendor's current HIPAA documentation and talk to your own legal or compliance adviser about your situation.*