ChatterLocks

A design for publishing AI chat transcripts with the private parts removed, in a way readers can check.

People who use AI to help write something increasingly want to show their work: what they asked, what the model said, where they pushed back. A methods section for AI-assisted writing. ChatterLocks lets an author publish a redacted transcript, prove that nothing visible was altered, and reveal a withheld passage later to anyone who challenges it.

The goal is modest. Showing your work should cost less than faking it, so a reader who dislikes slop, not AI, can tell the difference without taking anyone's word for it.

Status

This is a design document. Nothing is built, and nothing has been reviewed by a cryptographer. Do not rely on it for anything yet.

Read the specification (revision 0.11).

About this site

Plain static files, no analytics, no cookies, and nothing loaded from anywhere else. The host, Cloudflare, sees requests as any web host does; ChatterLocks keeps no logs of its own. If you need to read this unobserved, use Tor or an archived copy.