They See Your Photos

They See Your Photos is a privacy research experiment created by Ente that demonstrates how much private information modern AI image recognition can infer from a single uploaded photograph. By submitting an image to the Google Vision API, the tool reveals details about location, demographics, emotional state, economic status, and even lifestyle assumptions that AI systems can extract from visual data alone. The experiment serves as an eye-opening demonstration of the privacy implications of photo sharing in an era of powerful computer vision technology.

How It Works

Users upload a photo to the website, which transmits the image through Cloudflare to Google Vision API for analysis. The AI processes the image and generates a detailed natural language description covering the scene, subjects, environment, and inferred characteristics. The original uploaded image is immediately discarded after processing and is not stored on the server. All data transmissions are encrypted using TLS. The experiment also accesses EXIF location metadata when present and uses OpenStreetMap and Google Maps to provide location context.

What the AI Reveals

• Scene description: The AI identifies the setting, lighting conditions, time of day, and environmental context such as forests, urban areas, or indoor spaces.
• Subject profiling: The tool may infer approximate age, gender, ethnicity, emotional state, and economic status based on clothing, setting, and visual cues.
• Activity and intent analysis: The AI can determine whether a photo is candid or posed and describe the likely activity or purpose of the image.
• Image quality assessment: Details about camera type, resolution, potential digital editing, and photographic technique are included in the analysis.

Privacy Implications

The experiment highlights how much personal data is embedded in everyday photographs. Even without metadata, AI systems can extract meaningful information about individuals from visual content alone. Ente, the company behind the experiment, is an end-to-end encrypted photo storage service that positions itself as a privacy-focused alternative to Google Photos. The experiment underscores why encryption matters: while services like Google Photos can analyze uploaded images, Ente encrypts photos before they leave the user device, preventing server-side AI analysis.

Technical Details

Images are sent to Google Vision API via Cloudflare for AI processing. The tool uses Sentry for error monitoring and PostHog for anonymized product analytics. Location metadata lookups are powered by OpenStreetMap. No cookies are placed by the experiment itself, though third-party services may place their own cookies during service requests. The source code and infrastructure are managed by EnteIO Technologies Private Limited, based in Bengaluru, India.