Contents

  1. Introduction
  2. Privacy-by-Design Framework
  3. What We Collect
  4. Data Retention Schedule
  5. What We Never Do
  6. Volunteer Accountability
  7. AI Ethics & Bias
  8. Regulatory Alignment
  9. Contact & Requests

1. Introduction

Amber's Angels Inc. is a Georgia-based 501(c)(3) nonprofit organization that deploys a volunteer-driven, AI-powered search platform during government-issued missing person emergencies (AMBER Alerts, Silver Alerts, and others).

We are committed to the highest standards of privacy and ethical AI deployment. Our platform is designed to act as a force multiplier for law enforcement while strictly adhering to a Privacy-by-Design philosophy.

Plain language summary: We never store your video. We only run when an alert is active. We never sell your data. Every AI detection is reviewed by a human before anything reaches law enforcement.

2. Privacy-by-Design Framework

Unlike traditional surveillance systems, Amber's Angels does not build persistent databases of citizen movement. Our technology is event-triggered and ephemeral by architectural design — not just by policy.

No Video Archiving

Raw video streams from volunteer drones and smartphones are processed entirely in memory. Once the inference engine analyzes a frame, it is immediately discarded — it is never written to disk, transmitted to a server, or accessible to any human. This is enforced at the code level, not by an administrative policy that could be overridden.

Event-Triggered Only

Our detection pipeline only activates during declared, government-issued alerts consumed from the FEMA IPAWS official feed. Between alerts, the platform is completely dormant. There is no passive background collection.

Mission Specificity

Volunteers explicitly opt in to each individual search mission. Accepting one mission does not grant consent for any other. Volunteering for one alert type (e.g., AMBER) does not auto-enroll you in others.

Facial Obfuscation

Our processing pipeline includes automated protocols to obfuscate faces and the license plates of non-suspect vehicles in any data that progresses to coordinator review. Innocent bystanders are protected by default.

3. What We Collect

We practice strict data minimization — we collect only what is operationally necessary for the active mission.

From Volunteers (App Users)

Detection Events

Website Visitors

This website is hosted on GitHub Pages. We do not use cookies, tracking pixels, or analytics scripts. GitHub may collect standard server access logs (IP address, browser type, page requested) per their own privacy policy.

4. Data Retention Schedule

Data Type Purpose Retention Period Who Can Access
Raw video frames Real-time inference only Not retained — deleted post-inference Processing system only (ephemeral)
GPS telemetry Mission mapping and coverage analysis 90 days Mission coordinators only
Detection records Law enforcement leads (plate + VMMC data) 1 year Coordinators + law enforcement on active case
Volunteer identity Background checks, accountability, vetting Duration of volunteer status + 1 year Admin only
Alert metadata FEMA IPAWS source, timestamp, activation log 2 years (compliance) Admin only

All retention schedules are enforced at the infrastructure level through automated deletion routines — not dependent on manual processes.

5. What We Never Do

6. Volunteer Accountability & Vetting

All volunteers must complete a background screening prior to receiving platform access. This is a requirement of our law enforcement partnerships and a commitment to the communities we serve.

Vetting Requirements

Data Use Agreement

Volunteers explicitly agree to the following prohibitions: no independent video recording during missions, no social media posts about active searches, no sharing of operational details outside the platform, no contact with suspects or law enforcement on behalf of the organization. Violations result in immediate revocation of access.

7. AI Ethics & Bias Mitigation

Our AI models (YOLOv8, MobileNetV3, OpenALPR) are mission-scoped detection tools — not autonomous decision-makers.

Human-in-the-Loop by Design

No detection result is ever transmitted directly to law enforcement. Every potential match is reviewed by a trained coordinator before escalation. The AI surfaces leads; humans validate them. This is a structural constraint, not a preference.

Bias Awareness

We acknowledge that vehicle recognition AI can carry training data biases. Our team monitors for disparate false-positive rates across vehicle types and geographic contexts. We are committed to publishing model performance audits as our labeled dataset grows through pilot operations.

Transparency

Our general system architecture and data governance policies are public. We welcome scrutiny from civil liberties organizations, academic researchers, and grant oversight bodies. To request a technical briefing, email info@amberangels.org.

8. Regulatory Alignment

FrameworkOur Position
COPPANo minors may create volunteer accounts. The platform does not interact with child users.
CCPA / State Privacy LawsEphemeral processing and no commercial data sharing keeps us within applicable state-level requirements.
FAA Part 107All drone operations require certified pilots or supervised training. No autonomous drone flight.
FEMA IPAWSWe consume official IPAWS feeds only. We do not generate or amplify unofficial alerts.
GDPR (informational)Our privacy-by-design architecture aligns with GDPR principles, positioning us for future expansion.

9. Contact & Data Requests

Questions regarding this policy, privacy concerns, or requests for records deletion may be directed to:

Amber's Angels Inc.
Email: info@amberangels.org
Website: amberangels.org

We will respond to all privacy requests within 30 days.