AI Surveillance Research - AI Privacy Shield

Documented evidence of tracking, telemetry, and privacy violations by major AI platforms

These Companies Are Watching Everything

Our research documents how AI platforms collect keystroke data, session recordings, and behavioral telemetry - often ignoring user opt-out preferences.

🎯 GOTCHA: OpenAI

Fake favicon deception: Every URL you mention in ChatGPT gets sent to Google via disguised "favicon" requests. Your conversation topics = Google's advertising profile.

CRITICAL - URL LEAKS

🎯 GOTCHA: Anthropic

1-pixel hardware fingerprinting: Audio signatures, GPU hashes, and hardware UUIDs hidden inside "tracking pixels." Espionage-level surveillance disguised as images.

CRITICAL - PIXEL FINGERPRINTING

🎯 GOTCHA: Grok

Voice playback tracking: WebSocket parameter use_time_based_playback_tracking=true means every pause, replay, and speed change is logged. Cognitive surveillance via audio.

HIGH - VOICE TRACKING

🎯 GOTCHA: Gemini

WAA-PA fingerprinting hub: Google's central waa-pa.clients6.google.com endpoint captures audio fingerprints across all services. Gemini feeds Google's surveillance empire.

HIGH - AUDIO FINGERPRINTING

OpenAI / ChatGPT (Full Report)

Complete forensic analysis: Datadog RUM, keystroke telemetry, opt-out fraud, session recording, and 2,909 surveillance requests per hour.

CRITICAL - Full Analysis

Anthropic / Claude (Full Report)

Complete forensic analysis: Statsig analytics, Honeycomb tracing, Segment integration, and the Haiku model substitution scandal.

Full Analysis

Google / Gemini (Full Report)

Complete forensic analysis: Google advertising ecosystem integration, analytics cookies, and cross-platform tracking infrastructure.

Full Analysis

xAI / Grok (Full Report)

Complete forensic analysis: PII transmitted in plaintext, unencrypted telemetry, Mixpanel tracking queues, and aggressive data collection.

Full Analysis