Privacy by Design for Vulnerable Populations

The Privacy Imperative

Humanitarian operations often involve vulnerable populations:

• Refugees and displaced persons
• Survivors of conflict or disaster
• People in sensitive situations

Technology used in these contexts must protect, not expose.

Privacy Architecture

UPAS is designed around privacy:

No Data Exfiltration

Queries never leave the device:

• On-device inference
• Local cache only
• No server calls for queries

Minimised Logging

UPAS logs only:

• Cache health metrics
• Runtime status
• Sync timestamps

UPAS never logs:

• User queries
• Response content
• Operational context

No Tracking

UPAS does not:

• Set tracking cookies
• Collect analytics
• Fingerprint devices
• Store behaviour patterns

Implementation

Privacy is enforced architecturally:

// No analytics endpoints
// No telemetry calls
// No user identifiers

async function query(question) {
  // All processing local
  const result = await localInference(question);
  
  // No server call
  // No logging of question
  // No storage of context
  
  return result;
}

Data Sovereignty

Organisations control their data:

• Procedure packs hosted on their infrastructure
• No third-party data processing
• Complete visibility into data flows

Compliance

UPAS design supports:

• GDPR requirements
• Humanitarian data protection standards
• Organisational privacy policies

Trade-offs

Privacy constraints affect some features:

• No cloud-based model improvements
• No cross-device sync of preferences
• No usage-based optimisation

These trade-offs are intentional.

Learn More

• Security Model
• Deployment