Zenveus founder portrait
Zenveus founder portrait
Zenveus founder portrait
Zenveus founder portrait

Trusted by founders and incubator-backed teams

HIPAA for Lovable and Supabase Apps: What Has to Change

November 4, 2026 • 5 min read • Team Zenveus
Isometric illustration of a database cylinder inside a shield, with a padlock and a blue audit trail leading to a medical record card

Introduction

Lovable and Supabase are a fast way to get a healthcare idea in front of clinicians. The question arrives the moment a real patient record is involved: can this app hold protected health information (PHI), and what has to change first?

The short answer is that the prototype can become a HIPAA-ready product, but the PHI should not flow through Lovable, and Supabase only becomes an option once a Business Associate Agreement (BAA) and its HIPAA add-on are in place. Even then, most of the work sits in the app itself. This guide covers what each platform says, what a BAA does not cover, and the changes to make before the first real patient signs in.

This is an engineering guide, not legal advice. Your HIPAA obligations, risk analysis and agreements stay with your organization and its advisers.

What you’ll learn

  1. Whether you can put patient data into a Lovable app
  2. What Supabase requires before it can hold PHI
  3. What a BAA does not cover in an AI-built app
  4. The app-level changes HIPAA’s technical safeguards point to
  5. Every other service that needs a BAA, or no PHI at all
  6. A practical order of work before launch

Can you put patient data into a Lovable app?

Not on Lovable’s standard terms. Lovable’s terms ask users not to upload, input or otherwise provide PHI subject to HIPAA through its services unless their plan or a separate written agreement, such as a data processing addendum or enterprise agreement, expressly permits it (Lovable).

The same section says Lovable’s standard services are not designed to serve as a system of record for that data, or to provide regulatory-grade safeguards for it. In practice that splits the work in two. Lovable is where you design and build, using synthetic or de-identified data. Real patient data lives on infrastructure you control and have agreements for: your own Supabase project, your own hosting, and your own AI, email and analytics providers.

If the app runs on Lovable Cloud, the built-in backend Lovable provides on Supabase’s open-source foundation, that backend is part of Lovable’s services too, so check what your plan permits before treating it as a home for PHI. Lovable’s docs note that moving a project from the built-in backend to your own Supabase project is not a one-step change: you export and restore the database, move storage files and code, and set up authentication, secrets and external services again (Lovable Cloud). Plan that move before real patients arrive.

That split also covers the easy-to-miss paths. Pasting a real record into a prompt to debug a screen, or importing a production export to test a feature, sends PHI through the builder just as surely as a live form does.

What does Supabase require before it can hold PHI?

A signed BAA and the HIPAA add-on. Supabase’s documentation says organizations must have both when dealing with PHI (Supabase), and its BAA states that it is binding on Supabase only if the customer has subscribed to the HIPAA add-on, unless agreed otherwise in writing (Supabase BAA).

With the add-on enabled, each project that holds PHI is set to High Compliance in General Project Settings. Supabase then runs continuous checks and shows warnings in its Security Advisor when a setting does not meet the required configuration, which includes:

  • Point in Time Recovery enabled
  • SSL enforcement turned on
  • Network restrictions enabled
  • Postgres connection logging kept on

Two limits are worth knowing early. Supabase says these controls are not supported out of the box in self-hosted Supabase, and that HIPAA compliance there is out of scope for its documentation (Supabase). And its HIPAA guide describes a shared responsibility model: the platform has controls in place, while configuring projects and applying the recommended controls is the customer’s job.

What does a BAA not cover in an AI-built app?

The application. A BAA covers how a provider protects data on its platform. It does not check whether your row-level security lets one patient read another patient’s records, whether a secret key ships to the browser, or whether an Edge Function writes PHI into its logs.

Those are the gaps AI-built apps tend to have. In the production issues we reviewed across about 25 projects in October 2026, authorization gaps were among the most common patterns: identity trusted from the client, authorization defined but not enforced, and records reachable directly by ID. On Supabase, that usually shows up as one of these:

  • A table with row-level security turned off, or a policy that checks that a user is signed in but not that the row belongs to them.
  • The secret key (formerly the service_role key) used in client code, which bypasses row-level security entirely. Supabase is clear that the publishable key is public and only reaches what row-level security allows (Supabase API keys).
  • A public storage bucket holding uploads such as referral letters or lab results.
  • Error messages, analytics events or function logs that include names, dates of birth or notes.

Our free Row-Level Security Auditor checks the first of these in an afternoon, and the Fix Library covers common errors such as Supabase returning an empty array and storage uploads blocked by RLS.

Which app-level changes do HIPAA’s technical safeguards point to?

The HIPAA Security Rule’s technical safeguards cover access control, audit controls, integrity, person or entity authentication and transmission security (45 CFR 164.312). Each one maps to concrete work in a Lovable and Supabase app.

  • Access control means a unique account for every user, row-level security on every table that holds or links to PHI, scoped to the patient, clinician or clinic, and automatic sign-out after inactivity.
  • Audit controls mean a record of who viewed or changed which patient record and when, kept somewhere the app’s users cannot edit.
  • Integrity means PHI cannot be changed or deleted without that change being recorded, and backups you have actually restored from.
  • Authentication means verifying identity before access, with stronger sign-in for staff accounts that can see many patients.
  • Transmission security means encrypted connections everywhere, which Supabase’s SSL enforcement covers for the database, plus the same for every other service in the path.

Our HIPAA-aligned engineering checklist turns these into a list you can work through with your developer.

Which other services need a BAA, or no PHI at all?

Every service that receives PHI on your behalf. An AI-built health app usually sends data to more places than its builder realizes, so list them before launch:

  • The AI model provider behind summaries, transcription or chat
  • Transactional email and SMS
  • Error tracking and logging
  • Product analytics and session recording
  • File storage and document generation
  • Payments, scheduling and video tools

For each one, either sign a BAA or make sure no PHI reaches it. Often the second option is simpler: send a record ID to analytics instead of a name, strip notes from error reports, and keep message bodies generic (“You have a new message”) with the detail behind sign-in.

In what order should you make these changes?

Work from the data outward, so the riskiest gaps close first:

  1. Keep building in Lovable with synthetic or de-identified data only.
  2. Sign the Supabase BAA, enable the HIPAA add-on, and set the production project to High Compliance.
  3. Turn on row-level security for every table and write ownership-based policies, then test them with two accounts.
  4. Move every secret key to the server, and make storage buckets private with policies for each path.
  5. Add audit logging for patient-record access and changes.
  6. List every other service in the path, and sign a BAA or remove PHI from it.
  7. Restore a backup once, end to end, before real patients arrive.

None of this requires throwing away the prototype. In most Lovable apps the interface and workflows carry over, and the changes sit in the database policies, the server functions and the services around them.

Why is HIPAA a production readiness question?

The agreements make PHI allowed on a platform. The engineering decides whether it is protected in your app. Authorization and data isolation, secret management, and the compliance and audit surface are three of the nine checks in the Zenveus Production Readiness Standard, and they are where AI-built health apps most often fall short.

We have built healthcare software with this data in it: an AI clinical documentation platform for behavioral health with audio and transcript capture, specialist note categories and EMR integrations, designed for HIPAA-aligned data handling. Read the clinical documentation case study, or see our healthcare work.

If you want a senior engineer to check your app, the 48-Hour Verdict reviews your codebase against the Standard for $299 and returns a signed decision with a prioritized fix list. For a deeper review of PHI flows, roles, audit trails and integrations, the Healthcare Engineering Readiness Review is a fixed-scope audit. If the main worry is who can read which rows, start with the Supabase security audit.

NEXT STEP

Find out what stands between your app and real patient data

A named senior engineer reviews your Lovable and Supabase codebase against the Production Readiness Standard and signs a written decision in 48 hours, with the fixes in order.

Get the 48-Hour Verdict ($299)

Taking a Lovable app live more generally? Read Lovable to production: what it takes to ship to real users.

Need an engineering partner, not just developers?

Zenveus works with founders as a technical leadership layer across validation, architecture, MVP, launch, and scale.

FAQs

Frequently Asked Questions

Is Lovable HIPAA compliant?

Lovable's terms ask users not to provide PHI subject to HIPAA through its services unless their plan or a separate written agreement expressly permits it. Build with synthetic or de-identified data in Lovable, and keep real patient data on infrastructure covered by agreements.

Is Supabase HIPAA compliant?

Supabase signs a BAA for customers who subscribe to its HIPAA add-on. Projects holding PHI are set to High Compliance, with Point in Time Recovery, SSL enforcement, network restrictions and connection logging enabled. The application's own controls remain your responsibility.

Can I use Lovable Cloud for patient data?

Lovable Cloud is part of Lovable's services, so check what your plan permits first. Moving to your own Supabase project with a BAA is not a one-step change, so plan it before real patients arrive.

Do AI model providers need a BAA too?

If PHI is sent to them, yes. Any provider that receives PHI on your behalf, including AI model, email, analytics and error-tracking services, needs a BAA or must receive no PHI at all.

Does a BAA make my app HIPAA compliant?

No. A BAA covers how the provider protects data on its platform. Row-level security, key handling, audit logging and what your app sends to other services are still up to you.

Still have questions? Book a consultation.

Scroll to Top

Optional tracking stays off until you enable it. You can change your choice at any time.

Essential — always active
Security, requested features, and remembering these choices.

Read our cookie policy