OpenAI’s October 20 release notes introduce locked chats: selected conversations stay in an account but disappear from ordinary history, pinned chats and search, with access protected by an account PIN or supported device authentication. The announcement describes a gradual rollout to Free, Go, Plus and Pro. The accompanying guide lists additional managed plans, so readers should check their own account rather than assume immediate access everywhere.
For agent workflows, the exclusions matter. OpenAI says eligible cloud conversations in Chat or Work can be locked, but Codex and local Work conversations cannot. Chats associated with scheduled or background tasks are also excluded, and a locked chat cannot create or update a schedule. Locking does not create end-to-end encryption or change retention or model-improvement settings, and it does not erase information remembered beforehand. Existing shared links remain accessible for previously shared content. Resetting a forgotten account PIN permanently deletes the content protected only by that PIN. These are documented behaviors, not results from our own testing.
Editorial take: Treat this as a conversation-visibility control, not a way to put an entire ongoing workflow behind a new privacy boundary. Before relying on it, identify which chat, files, shared copies and scheduled work actually belong together. A feature that hides a transcript may be useful during screen-sharing without replacing the separate decisions about who can access its outputs or whether recurring work should continue. The sources do not establish a new locking capability for Dots themselves.