Work
Veil — Privacy-Preserving Computer Vision
What Is Veil?
Veil is a privacy-first computer vision system that gives cameras the ability to see without identifying. Built for AI wearables like Meta Ray-Ban glasses, Veil automatically detects and obscures the identities of every person and animal in frame before any image or video is ever written to storage.
Tagline: See everything. Identify no one.
Status: Patent pending, working prototype. Meta AI Catalyst Grant application ($200,000) under review.
The Core Problem Veil Solves
AI wearable cameras are powerful but socially toxic — people don't want to be filmed without consent. Veil removes that barrier entirely. It's not a policy fix or a terms of service. It's architectural: identities are destroyed on-device before storage. There is no footage to leak, subpoena, or misuse.
How It Works
Veil is trigger-based. The live view is completely untouched — the wearer sees normally. The moment they record or capture a photo, Veil activates, runs on-device ML inference, detects every human and animal in frame, and obscures their identities before the image is written to storage. Zero data ever leaves the device.
The current prototype runs entirely on a Raspberry Pi Zero 2W — a $12 microcomputer — with no external dependencies. It is designed to port to Meta's Snapdragon-based hardware.
Key Features
- Trigger-based activation — only runs during recording, never on the live view
- On-device inference — zero cloud transmission, identities obscured before storage
- Dual subject detection — obscures both humans and animals
- Custom ML model optimized for minimal computational footprint
- Hardware-agnostic architecture — designed to port across platforms
- Patent pending — USPTO provisional patent filed
Phase 2 — Veil ID Consent Network
Phase 2 inverts the privacy model entirely. Instead of blanket obscuring, it introduces affirmative, hardware-authenticated consent. A subject taps their phone or a Veil ID tag to the wearer's glasses, generating a time-scoped consent token. That person is then visible in recordings; everyone else remains protected. Consent is time-limited and revocable at any time.
Use cases unlocked by consent layer: Journalism, content creation, family events, medical and clinical documentation (HIPAA-compliant by design), enterprise environments.
Tech Stack
- Custom ML model for human/animal detection
- On-device inference (Raspberry Pi Zero 2W prototype)
- Designed for Meta's Snapdragon-based hardware
- NFC/tap-based consent tokens (Phase 2)
Why Veil Matters
- Removes the primary social and ethical barrier to everyday AI wearable use
- Protects vulnerable individuals: domestic violence survivors, activists, whistleblowers, minors
- Positions wearable cameras for healthcare, education, and enterprise markets previously blocked by privacy concerns
- Open SDK vision for third-party developers to build privacy-first applications
Common questions
What is Veil? Veil is Blacksky's privacy-preserving computer vision system for wearable cameras. It obscures people's identities on-device, before footage is ever saved — so the camera sees everything but identifies no one.
Does Veil use the cloud? No. All processing happens on-device. Nothing is ever transmitted externally. The identity obscuring happens before storage, so there is no identifiable footage anywhere in the pipeline.
What devices does it work on? The current prototype runs on a $12 Raspberry Pi Zero 2W. Veil is being developed for Meta Ray-Ban glasses (Snapdragon hardware) with a Meta AI Catalyst Grant application currently under review.
Is Veil available yet? It's a working prototype with a patent pending. The full product launch is tied to the Meta grant decision (May 2026). Contact mario@blackskymedia.org for early access or partnership discussions.
Who is Veil for? Anyone using wearable cameras who needs to protect the privacy of bystanders — journalists, content creators, healthcare providers, parents at family events, enterprise teams. Phase 2 adds a consent layer for scenarios where some subjects want to be visible.