Asellera Privacy Controls Summary

1. PURPOSE

This document summarizes how privacy controls are applied across the Asellera data flow:

Patient → Communication Layer → AI Processing Layer → Workflow Engine → Data Storage → Clinic Systems → Human Review

At each stage, Asellera enforces technical, administrative, and procedural controls to protect personal information and personal health information (PHI).

2. Stage-by-Stage Privacy Controls

2.1 Patient → Communication Layer

Components:

• Voice calls (Retell AI or equivalent)

• SMS messaging providers

• Web intake forms

Privacy Controls:

• Consent prompts for voice/SMS interactions

• Minimum necessary data collection (only what is needed for task)

• TLS encryption for all inbound/outbound communication

• No unnecessary persistent storage at edge systems

• Phone/SMS masking where applicable

Protected Data:

• Phone numbers

• Message content

• Voice recordings (if enabled)

2.2 Communication Layer → AI Processing Layer

Components:

• Retell AI transcription systems

• SMS message ingestion

• API gateways

Privacy Controls:

• Secure API authentication (API keys / OAuth)

• Data transmitted over encrypted channels (TLS 1.2+)

• Short-lived processing of raw audio/text where possible

• No retention of raw audio beyond configured retention rules

• Access logging of all API calls

Key Principle:

Only necessary conversation data is forwarded to AI systems.

2.3 AI Processing Layer

Components:

• OpenAI or equivalent LLM systems

Privacy Controls:

• Data minimization before prompt construction

• No training of public models on PHI without explicit agreement

• Prompt and response logging for auditability

• Guardrails to prevent sensitive data overexposure in outputs

• AI restricted to assistive functions only (no clinical decision-making)

Protected Data:

• Transcripts

• Intake data

• Draft SOAP notes

• Patient context

2.4 AI Processing → Workflow Engine

Components:

• Asellera backend services

• n8n automation workflows

Privacy Controls:

• Structured data transformation (removal of unnecessary raw text where possible)

• Role-based workflow execution

• Secrets management for integrations

• Limited access to full conversation history

• Logging of workflow triggers (not full PHI where avoidable)

2.5 Workflow Engine → Data Storage Layer

Components:

• Supabase database

• Encrypted object storage (if used)

• Audit log systems

Privacy Controls:

• Row-Level Security (RLS) enforcement

• Role-based access control (RBAC)

• Field-level data access restrictions (where applicable)

• Encryption at rest (AES-256 or equivalent)

• Structured separation of:

  • identity data

  • clinical data

  • system logs

• Retention policies enforced at database level

2.6 Data Storage → Clinic Systems

Components:

• PMS systems

• EHR integrations

• Clinic dashboards

• Google Calendar

Privacy Controls:

• API authentication for all outbound data sync

• Data minimization before export (only required fields sent)

• Secure transmission (TLS encryption)

• Access scoped per clinic tenant (multi-tenancy isolation)

• No cross-tenant data leakage possible by design

2.7 Clinic Systems → Human Review Layer

Components:

• Healthcare providers

• Clinic staff

• Admin users

Privacy Controls:

• Role-based access controls inside clinic accounts

• Audit logs of all data views and edits

• Separation of draft vs finalized clinical outputs

• Human approval required for all AI-generated clinical documentation

• Access restricted to assigned clinic only (tenant isolation)

3. Cross-Cutting Privacy Controls (All Stages)

3.1 Encryption

• TLS 1.2+ for all data in transit

• AES-256 (or equivalent) for data at rest

• Encrypted backups

3.2 Access Control

• Role-Based Access Control (RBAC)

• Multi-Factor Authentication (MFA) for sensitive systems

• Principle of least privilege enforced

• Unique user identities (no shared accounts)

3.3 Data Minimization

• Only necessary patient data is collected

• AI prompts are constructed with minimal required context

• Non-essential identifiers are excluded where possible

3.4 Audit Logging

Logged events include:

• Data access events

• Authentication attempts

• API calls involving PHI

• Workflow executions

• AI generation events

Logs are used for:

• SOC 2 audits

• Incident response

• Compliance monitoring

3.5 Tenant Isolation

Each clinic operates in a fully isolated environment:

• No cross-clinic data access

• Separate logical partitions in database

• Scoped API keys per tenant

• Isolated workflow execution contexts

3.6 AI Privacy Controls

• AI cannot independently store or retain memory beyond system design

• No autonomous clinical decisions

• Outputs are treated as draft suggestions only

• Sensitive outputs are logged securely, not exposed broadly

4.Data Retention Enforcement

Privacy controls enforce retention rules from:

• Data Retention Policy

• Clinic-specific configurations

• Legal/regulatory requirements

Automated deletion and archival mechanisms ensure compliance.

5. Vendor Privacy Controls

All vendors processing PHI must:

• Sign data processing agreements (DPAs)

• Enforce encryption in transit and at rest

• Restrict use of data to service delivery only

• Support deletion requests where applicable

Key vendors include:

• OpenAI

• Amazon Web Services

• Retell AI (voice layer)

• Supabase (database layer)

• Messaging and scheduling providers

6. Incident Privacy Safeguards

In the event of a security incident:

• Affected data is immediately isolated

• Access is revoked or restricted

• Logs are preserved for forensic review

• Breach impact assessment is performed

• Regulatory obligations are evaluated

7. Summary of Privacy Model

Asellera privacy is enforced through:

• Technical controls (encryption, RBAC, tenant isolation)

• Administrative controls (policies, audits, approvals)

• Operational controls (logging, monitoring, incident response)

At every stage of the system:

Data is minimized, protected, isolated, and auditable.

Approval

Approved By: Shane Senha, CEO
Company: Asellera
Version: 1.0
Date: June 16th 2026