Security & Privacy

Last updated: May 27, 2026

Ada is a personal health record app. Everything we build is in service of one promise: that the information you trust us with stays protected, stays yours, and is used only for the purposes you authorize. This page summarizes how we protect that information, the technical safeguards in place, and how healthcare organizations, security researchers, and patients can reach our security team.

For hospital security teams: we welcome integration reviews. Email [email protected] for our SIG (Standardized Information Gathering) responses, BAA template, or to request our public JWK for SMART-on-FHIR client authentication.

The short version

Encryption in transit

TLS 1.2+ everywhere

All traffic between the Ada app, our API, and partner systems is encrypted in transit. We pin Apple App Attest certificates on the iOS client.

Encryption at rest

AES-256, per-record

Identifiable health records are stored with AES-256 encryption at rest. Field-level encryption for sensitive content like clinical notes.

Access controls

Per-user data isolation

No employee has standing access to identifiable patient data. Production access is logged, time-bounded, and requires documented justification.

Audit logging

Append-only, immutable

Every access to a patient's record is logged with actor, timestamp, action, and resource. Logs are retained per HIPAA guidance.

Device attestation

Apple App Attest

API requests from the iOS app are cryptographically attested. We verify the request came from the genuine Ada app on a real Apple device.

Infrastructure

AWS US-East, hardened

Production runs on AWS in HIPAA-eligible services. No SSH; all admin access is via AWS Systems Manager (SSM) session logs.

1. Data minimization

We collect only what's needed to deliver the Service: account information (name, email, date of birth), records you upload or import at your direction, and limited device/usage telemetry. We do not collect precise GPS location, advertising identifiers, or social-graph data. We do not sell information that identifies you, and we do not use your identifiable health information for advertising. (See Privacy Policy.)

2. Encryption

In transit

At rest

Key management

3. Authentication & access

End-user authentication

EHR integrations

Administrative access

4. Audit logging & monitoring

5. AI processing

Ada uses Anthropic's Claude as a sub-processor for record summarization and conversational features. Inputs and outputs are processed under Anthropic's Commercial Terms of Service, which restrict use of customer data to delivering the service (no training on customer data). We retain only the minimum necessary AI conversation history to deliver continuity for the user, and that history is encrypted at rest. Patients may opt out of AI features per record. See Privacy Policy §7 for the full AI processing disclosure.

6. Subprocessors

We use a minimal set of vendors. Current production subprocessors handling identifiable information:

We will notify customers and update this page before introducing a new subprocessor that handles identifiable health information.

7. Breach response

8. Patient rights & deletion

9. Compliance posture

10. Responsible disclosure

If you believe you have found a vulnerability in Ada, please email [email protected]. We commit to:

We are working on a formal bug-bounty program; in the meantime we can offer acknowledgment and, where appropriate, a thank-you payment.

Contacts

Security inquiries / vuln reports: [email protected]

Hospital integrations / BAA: [email protected]

Privacy requests: [email protected]

Legal: [email protected]

Public JWK for SMART client authentication: /keys/ada-public-jwk.json · /.well-known/jwks.json