On AWS the term is HIPAA eligible, not HIPAA compliant, and the difference decides how you build. AWS publishes a list of services that are eligible to create, receive, process, maintain or transmit electronic protected health information. Eligibility means AWS permits that service to handle PHI once you have accepted the AWS Business Associate Addendum. It does not mean the service is compliant on your behalf. AWS says so directly: customers must still configure these services consistent with HIPAA requirements, and customers are responsible for ensuring their own HIPAA compliance. No AWS service makes you HIPAA compliant by itself.
Eligible is not compliant.
Every row below is taken from the AWS HIPAA Eligible Services Reference, which AWS last updated on 3 August 2026. This is a working subset, not the full list, which runs to well over a hundred services. Always check the live AWS page before an architecture decision, because AWS adds services to it over time.
| AWS service | Eligible under a BAA | What still needs configuring |
|---|---|---|
| Amazon RDS | Eligible | SQL Server, MySQL, Oracle, PostgreSQL, Db2 and MariaDB engines only. Encrypt at rest with KMS, force TLS in transit, and remember snapshots inherit the encryption decision. |
| Amazon Aurora | Eligible | Both the MySQL and PostgreSQL compatible editions. Storage encryption is set at cluster creation and cannot be toggled later. |
| AWS Lambda | Eligible | Environment variables are not a secret store. Use Secrets Manager or KMS, run in a VPC when reaching private data, and keep PHI out of logs. |
| Amazon S3 | Eligible | Default encryption on, Block Public Access on, versioning and access logging enabled. Most PHI exposure stories start with a bucket policy. |
| Amazon EC2 | Eligible | Encrypt EBS volumes and snapshots, patch to a baseline, and keep instances holding PHI in private subnets. |
| Amazon ECS | Eligible | Task definitions should pull secrets by reference, not literal values. Container logs are a common accidental PHI sink. |
| Amazon EKS | Eligible | Encrypt secrets at rest with KMS, keep the API endpoint private where possible, and enable audit logging. |
| AWS Fargate | Eligible | ECS and EKS engines only, per the AWS reference. Same logging and secrets discipline as the orchestrator above it. |
| Amazon DynamoDB | Eligible | Encryption at rest is on by default. Use a customer managed key where you need an independent authorization gate. |
| Amazon ElastiCache | Eligible | Enable encryption in transit and at rest explicitly. Caches holding PHI are easy to forget in an evidence pack. |
| Amazon CloudFront | Eligible | Excludes content delivery through CloudFront Embedded Points of Presence. Check this before assuming full coverage. |
| Amazon SES | Eligible | Email carrying PHI needs enforced TLS and careful handling of bounce and complaint data. |
| Amazon SNS | Eligible | Encrypt topics with KMS. Message bodies and delivery logs both count as places PHI can leak. |
| AWS KMS | Eligible | The key layer under most of the above. Scope key policies so a misconfigured IAM policy alone is not enough to decrypt. |
| Amazon CloudWatch Logs | Eligible | Eligible, but treat it as a PHI store if your application logs PHI. Encrypt log groups and set retention deliberately. |
Source: AWS HIPAA Eligible Services Reference, the canonical list. One point people miss: a service not on that list is not banned from your account. AWS allows you to use services that are not listed inside a HIPAA account, provided they do not process or store PHI. The boundary is where PHI goes, not which services exist.
RDS is HIPAA eligible, not HIPAA compliant, and the AWS reference qualifies it: SQL Server, MySQL, Oracle, PostgreSQL, Db2 and MariaDB engines only. You can store PHI in RDS once your BAA is accepted. What you still owe: encryption at rest with KMS enabled at creation time, TLS enforced for connections rather than merely available, and automated backups and snapshots treated as PHI themselves, because an unencrypted snapshot is exactly the gap an auditor looks for. Enable enhanced logging and ship it somewhere the database team cannot quietly edit.
Lambda is HIPAA eligible under a BAA. The configuration essentials are specific to how people actually use it: keep credentials in Secrets Manager or KMS rather than environment variables, which are readable by anyone with function read access; run functions in a VPC when they reach private data so PHI does not traverse the public internet; and audit what you log, because a debug line printing a request body is the most common way PHI ends up somewhere nobody classified as a PHI store.
Aurora is HIPAA eligible under a BAA, and AWS confirms both the MySQL compatible and PostgreSQL compatible editions. If your BAA is already accepted, no further step is needed to start using it in the covered accounts. The configuration catch worth knowing early: Aurora storage encryption is chosen when the cluster is created and cannot be switched on afterwards, so an unencrypted cluster means a migration rather than a setting change. Pair that with TLS in transit, IAM database authentication where it fits, and audit logging retained long enough to answer an investigation.
Under HIPAA, a cloud provider handling PHI is a business associate. The Business Associate Addendum is the contract that makes that relationship explicit and limits how AWS may use and disclose PHI.
AWS offers a standard BAA and you accept it yourself through AWS Artifact in the console, for a single account or across an AWS Organization. It is self-service, so there is no sales process and nothing to negotiate.
You may use any AWS service in an account covered by the BAA, but PHI must only be processed, stored and transmitted in the eligible services. Account separation is the cleanest way to keep that boundary provable.
A SaaS provider signs the BAA with AWS, and each healthcare customer signs a BAA with the provider rather than with AWS. A covered entity that is also a direct AWS customer may need both.
Picking eligible services is the easy part and takes an afternoon. The work that decides whether an audit goes well is everything after it.
Our HIPAA on AWS page covers the full picture: BAA, encryption, access controls, audit logging and PHI segmentation, and how those become the evidence your auditors and compliance team need in order to attest.
A cloud security audit tells you where the environment actually stands against these controls, and AWS security hardening is the remediation work that closes what it finds and documents each change.
If you are a digital health or clinical SaaS team, our HealthTech on AWS page covers the wider picture, including GDPR Article 9 and NHS DSP alongside HIPAA. Our published UK HealthTech case study covers a 38% AWS cost reduction in 90 days on that kind of estate.
This page is engineering guidance, not legal advice. Your auditors and compliance team attest to HIPAA compliance; we build and evidence the infrastructure that lets them.
Bring your architecture and we will walk the PHI path through it with you: which services touch it, which are eligible, and where the configuration falls short of what an auditor will ask for. You get at least one concrete recommendation, yours to keep whether you engage us or not.
★ AWS Advanced Tier Services Partner · AWS-certified engineers · Founder on every engagement call