Is Claude HIPAA compliant?

By Mat Steinlin, Head of Information Security

Last updated: July 2026

Claude is HIPAA compliant, but with notable conditions. Claude for Enterprise, obtained through Anthropic's sales team, can be configured for HIPAA-compliant usage. Consumer Claude.ai products have no BAA coverage. Claude Code is eligible only with Zero Data Retention enabled, and several beta features have no coverage path at all.

This is the short version. For the full breakdown (coverage table, common failure modes, BAA process timeline, and what you need to build on top of it) see Is Claude HIPAA Compliant? BAA, Coverage, and Gaps.

Coverage at a glance

The most common mistake is assuming Claude.ai is covered because Anthropic has a BAA. Claude.ai is a consumer product. Claude for Enterprise is a separate, sales-negotiated offering.

How to get a BAA with Anthropic

How you get a BAA depends on how you're using Claude. For Claude for Enterprise, admins can sign the BAA directly in their settings when activating HIPAA compliance. For API access or self-managed accounts, contact Anthropic sales. See Anthropic's BAA documentation for the full eligibility breakdown.

For Claude Code specifically, Zero Data Retention (ZDR) must be enabled before the tool is eligible for HIPAA use. ZDR isn't included in the standard Enterprise plan and can't be toggled in admin settings. It requires separate approval from Anthropic and is only available to qualified accounts. Contact your Anthropic account team or reach out to sales to confirm eligibility.

Anthropic's ZDR retains inputs in RAM until the cache expires, rather than disabling the cache entirely. That distinction matters for highly sensitive PHI workflows where in-memory retention is still a concern.

On the BAA process itself: For Claude Enterprise, admins can sign the BAA directly in admin settings when activating HIPAA compliance; no sales call required. For API and self-managed accounts, you submit a form and wait for a sales response. Multiple practitioners have reported multi-week or multi-month wait times with no reply, or receiving only automated responses. The API path remains a meaningful friction difference compared to AWS or Azure, where HIPAA agreements are self-service.

If you want to avoid the Enterprise sales process entirely, Claude on AWS Bedrock offers a self-service HIPAA path through the standard AWS BAA. Practitioners increasingly recommend this route for teams that need HIPAA coverage without negotiating a direct Anthropic agreement. See Claude on Amazon Bedrock for details.

On pricing: HIPAA readiness is one of only a few genuine differentiators of Claude Enterprise over Team, but it comes at a price. Teams on the Claude Team plan hit a 150-seat cap and must move to Enterprise at that point, which adds usage-based pricing on top of seat fees. If your team is smaller and looking for a self-service HIPAA path, there is no off-the-shelf option on the Team plan; Enterprise negotiation is the only route Anthropic currently offers.

What the BAA requires of you

The BAA is not just a vendor obligation; it imposes contractual requirements on you. Anthropic's BAA (Section D.7) requires compliance with their Implementation Guide for HIPAA Entities, and data submitted in a manner inconsistent with that Guide is not treated as PHI under the BAA. Key requirements include: workforce training before provisioning access, prohibited use restrictions (Claude may not be used for direct patient-provider interactions or to diagnose patients), authentication and access controls (SSO, MFA, lockouts, automatic logoffs), and separate BAAs for any third-party integrations including MCP servers and connectors.

For the full breakdown of Implementation Guide requirements, see Is Claude HIPAA Compliant? BAA, Coverage, and Gaps.

What the BAA doesn't cover

Most HIPAA failures with LLM usage happen not at the provider, but in the layer your team owns. A signed Anthropic BAA doesn't log your prompts, encrypt your audit trail, or enforce access controls on your API keys. See What an AI BAA actually covers for the full breakdown.

For teams who don't want to build that layer themselves, Aptible LLM Gateway provides it as a managed compliance layer, and your Aptible BAA covers all supported models, including Claude.

Full details

For the complete guide (including the coverage table, a breakdown of the ZDR requirement, the shadow AI risk for developer workflows, and the FAQ) see the dedicated chapter:

→ Is Claude HIPAA Compliant? BAA, Coverage, and Gaps

What's next

For teams using multiple AI vendors, or looking to manage the full compliance layer across models: Managing BAAs across multiple AI vendors.

For the infrastructure requirements that apply once you have a BAA in place, see What an AI BAA actually covers.

__

Anthropic's BAA covers their infrastructure. The audit logging and key management that HIPAA requires on top of it are still yours to build. And if you're using Claude alongside other models, you're managing separate BAAs for each vendor.

Aptible LLM Gateway handles the compliance layer as a managed service: every prompt and response logged and encrypted automatically, PHI scrubbed before it reaches the model, and one Aptible BAA covering Claude and every other supported model. Talk to an engineer to see if it fits your stack.