No AI tool is HIPAA compliant on its own. Compliance comes from the agreement, the configuration and how your people use it. Anthropic signs a business associate agreement for HIPAA-ready Claude Enterprise plans and for its API, but only some features are covered, the setup can't be reversed, and personal plans have no BAA at all. PHI belongs only in covered features.
"Is it HIPAA compliant?" is the question every health organization asks about AI. The better question is: which plan, which features, and what agreement?
Why does a BAA matter?
Because HIPAA lets a covered entity share PHI with a vendor only with "satisfactory assurance" that the vendor will safeguard it, documented in a written business associate agreement. An AI tool that processes PHI without one is a disclosure you can't defend. That rules out personal accounts immediately, free or paid.
Which Claude plans come with a BAA?
As of October 2026, per Anthropic:
| Plan | BAA available? | What to know |
|---|---|---|
| Free, Pro, Max (personal) | No | Consumer terms; never for PHI |
| Team | Not listed as covered on Anthropic's BAA page | Don't use it for PHI |
| Enterprise | Yes, once the Primary Owner turns on the HIPAA-ready setup and accepts the BAA | One-way decision; covers only some features |
| API (first-party) | Yes, for HIPAA-ready services | Enabled with Anthropic; some API features aren't covered |
A standard Enterprise plan doesn't include BAA coverage until the Primary Owner acts. The BAA covers only the organization that accepted it.
What does the Enterprise BAA cover, and what doesn't it?
| Covered | Available but not covered | Not available with the HIPAA setup |
|---|---|---|
| Chat, Projects, Artifacts, Voice, Web Search, Research, file creation and code execution (without network access) | Skills, connectors (MCP), Enterprise Search, Claude in Chrome, some Microsoft 365 features | Claude Design, Slides and Docs (beta), and several cloud and remote features |
For Claude Code, terminal sessions are covered only with the HIPAA configuration applied, signed in with the Enterprise account and connected directly to Anthropic. The VS Code and JetBrains extensions are available but not covered, and sessions routed through other clouds or gateways aren't covered.
The practical rule: PHI goes only into covered features. Turn off, or restrict by role, everything in the middle column for anyone who works with PHI.
What does Texas add?
Two state laws matter for Texas providers using AI:
- SB 1188 (effective September 1, 2025): a practitioner may use AI for diagnostic purposes, including recommendations on diagnosis or treatment, if they review all AI-created records under Texas Medical Board standards and disclose that use to patients. It also requires electronic health records with patient information to be physically maintained in the United States, including records held by cloud providers, for storage on or after January 1, 2026.
- TRAIGA (effective January 1, 2026): if AI is used in relation to a health care service or treatment, the provider must disclose it to the patient or their representative no later than the date the service is first provided, or as soon as reasonably possible in an emergency.
How each applies to your specific use, and where AI records and logs live, is a question for counsel. Ask your vendor in writing where data is stored and processed.
What does Anthropic itself require?
Anthropic's Usage Policy, taking effect November 12, 2026, treats medical recommendations as high-risk. A qualified person must meaningfully review any AI recommendation on diagnosis, treatment or medication before it reaches the patient, and the patient must be clearly told AI was used. It applies to anyone using Anthropic's products for those recommendations, not just consumer apps.
How should a health organization set it up?
- Enterprise only, HIPAA-ready, BAA accepted before any PHI goes in.
- Restrict uncovered features for anyone who handles PHI: connectors, Skills, Enterprise Search, Claude in Chrome.
- Keep a clinician in the loop on anything diagnostic or therapeutic, and document the review.
- Disclose AI use to patients where SB 1188, TRAIGA or Anthropic's policy calls for it.
- Minimize and de-identify where you can. Removing the name isn't de-identification; HIPAA Safe Harbor removes 18 categories of identifiers.
- Set retention and logging so PHI isn't kept longer than your policy allows.
An honest caveat: this is not legal advice. Coverage lists change, and Anthropic updates its BAA page often, so check it again before you go live.
How do we handle it at Andes?
We can support hospitals, clinics and health plans: we build systems for regulated health data, with compliance configured with the client and its counsel. We work with Claude every day and roll it out for companies. For the full configuration, see the Claude Enterprise setup checklist, and for what can go in at all, what data employees can put into AI tools.
Is your PHI only touching covered features?
If nobody has mapped that yet, start there. See our AI for regulated institutions service and let's review your setup before PHI goes in.
Sources
- 45 CFR 164.502, Uses and disclosures of protected health information (Cornell LII): https://www.law.cornell.edu/cfr/text/45/164.502
- Claude Help Center, Business Associate Agreements (BAA) for Commercial Customers: https://support.claude.com/en/articles/8114513
- Claude Help Center, HIPAA-ready Enterprise plans: https://support.claude.com/en/articles/13296973
- Texas Legislature, SB 1188 (89th Legislature, enrolled): https://capitol.texas.gov/tlodocs/89R/billtext/html/SB01188F.htm
- Texas Legislature, HB 149, Texas Responsible Artificial Intelligence Governance Act (enrolled), Sec. 552.051: https://capitol.texas.gov/tlodocs/89R/billtext/html/HB00149F.htm
- Anthropic, Usage Policy (effective November 12, 2026): https://www.anthropic.com/aup
- 45 CFR 164.514, De-identification (Cornell LII): https://www.law.cornell.edu/cfr/text/45/164.514