A new website, TheySeeYourPhotos.com, is turning heads by showing just how much Google’s AI can infer from your pictures.
Developed by the team behind the privacy-focused photo storage service Ente, the platform highlights how Google’s computer vision models process images, offering detailed, sometimes unsettling analyses of uploaded photos.
What the Website Does
Users can upload a photo to TheySeeYourPhotos.com, where Google Cloud’s computer vision AI generates a detailed, three-paragraph description of it. For example, the AI can identify clothing styles, weather conditions, and even approximate the time shown on a watch based on metadata.
However, concerns arose when the AI made inaccurate or sensitive assumptions, prompting Ente to adjust its prompts for more neutral outputs.
For those hesitant to upload personal photos, the site also includes stock images to test the AI’s capabilities. The goal is to showcase how much data AI models can extract from a single image—and to make users rethink storing their photos on platforms like Google Photos.
Privacy Concerns and Ente’s Solution
Vishnu Mohandas, the founder of Ente, launched the service to provide a secure alternative to Google Photos. Ente is open-source, end-to-end encrypted, and focused on privacy. Mohandas, a former Google engineer, stopped using Google Photos in 2020 after learning about the company’s involvement in AI-driven military projects.
He emphasizes that while Google Photos offers unmatched convenience, it’s also a treasure trove of personal data that could be used unpredictably in the future. “Photos taken today could reveal private details decades from now,” he warns.
Challenges Ahead
Switching to Ente isn’t seamless—Google Photos’ file structures and compression make data migration difficult. Additionally, Ente’s smaller scale means fewer advanced features like search and sharing. However, Mohandas believes in providing a viable option for those prioritizing privacy over convenience.
By creating TheySeeYourPhotos.com, Ente hopes to raise awareness about the potential misuse of visual data and encourage users to explore alternatives to mainstream platforms.
