Secure Render

Stage 1.

Stage 2.

SecureRender is a privacy protecting AI graphics runtime. It uses offline sandboxed WebWorkers & WebGL/GPU to run untrusted code. A ServiceWorker bricks itself into a polyfilled browser extension to emulate how SecureRender might work if it were implemented as a web standard.

SecureRender apps (games, edge functions, AI models, etc.) can instantly render and compute against local user data, this is done inside an offline enclave so the data remains local and cannot be accessed by the app's server, origin, or security context. Anything can be loaded into it, but nothing can come out.

As of June 2023 the prototype works as a proof of concept in desktop Chrome. Apps from one website can instantly render data from any other website using SecureRender, no login or registration needed. Models (mesh or AI) downloaded from one app would also immediately be available for use by any other site without needing to be reloaded. This feature may not be available in Brave or Firefox without command line flags or without the browser extension.

Since both models and user data are local and remain offline, any app can auto prompt inject entire user history and behavior into inferencing to instantly personalize AI agents, privately, without prompts ever being sent to a server API or a user needing to login or register. Each next app can do this, whether new or used before, building ontop of each other.


For more information, check out these links and references: