ChatGPT Group Chats Are Being Retired After a Global Pilot

ChatGPT group chats are being retired, not prepared for a wider launch. Under OpenAI’s retirement timetable, users have been unable to create a group chat, convert an individual conversation or join through an invitation link since July 9, 2026; existing rooms remain accessible while they transition to read-only status.

The discovery that appeared to expose a hidden feature in 2025 did lead to a released product, but not to a durable Slack competitor. Group chats progressed from a regional pilot to worldwide availability before OpenAI reversed course, leaving users with preserved conversation histories rather than an expanding collaboration platform.

What remains available during the wind-down

Existing participants may continue posting until their particular room becomes read-only, and the timing can vary during the transition. After that change, members can review messages, files and images already shared in the room but cannot add new material.

The retirement does not affect ordinary individual ChatGPT conversations. Users also do not need to take action simply to retain their group history, although anyone who wants a separate copy can request an account-data export through ChatGPT or the OpenAI Privacy Portal.

A group still using an active room must move its continuing discussion elsewhere before posting is disabled. Old invitation links cannot add replacement or additional participants, even while existing members can still exchange messages.

How the feature moved from rumor to release

OpenAI’s launch account dates the initial pilot to November 13, 2025, in Japan, New Zealand, South Korea and Taiwan, records the November 20 expansion to logged-in Free, Go, Plus and Pro users worldwide, and describes invitation-based rooms for as many as 20 people. The product therefore advanced well beyond traces found in web code: it became a consumer feature on both mobile and web.

A shared room placed human participants and ChatGPT in one conversation. When someone added people to an existing private thread, ChatGPT created a group copy and left the original conversation unchanged, preventing earlier private discussion from being converted in place.

TechCrunch’s contemporaneous coverage independently documented the global rollout, availability across the four consumer plan categories, the 20-person ceiling and the copied-thread behavior. Those details settled the original uncertainty over whether group chats were merely an internal experiment or a public pilot.

What group chats actually offered

The feature was designed for bounded collaborative tasks rather than continuous organization-wide communication. A room could support activities such as planning a trip, drafting an outline, comparing options or researching a topic while every participant worked from the same messages and shared material.

ChatGPT followed the human exchange and decided when to respond instead of answering every message automatically. Participants could mention ChatGPT when they wanted its input, while room-specific instructions allowed the group to set a common context or preferred response style.

The rooms also supported search, file and image uploads, image generation, dictation, emoji reactions and inline replies. These tools made the experiment more substantial than a simple link-sharing layer, but they still operated inside an individual conversation rather than across a company’s entire communications structure.

The privacy boundary kept accounts separate

Sharing one AI-assisted context did not merge the participants’ ChatGPT accounts. Personal memory was not used inside a group chat, and the conversation did not create new entries in an individual user’s memory.

That separation reduced the chance that information retained from an unrelated private conversation would appear automatically in front of coworkers, relatives or friends. It did not make material posted to the room private from its members: participants could see the shared conversation, uploaded content and membership.

Invitation links created a different privacy consideration. Any member could redistribute a valid link while joining remained enabled, so practical control depended partly on participants limiting where invitations travelled. A recipient still had to accept the invitation, but accepted members could view the conversation available to the group.

Why the Slack comparison went too far

Placing an AI assistant inside a conversation among several people invited comparisons with workplace messengers. The overlap was real but narrow: both formats supported discussion and shared material, while ChatGPT could also summarize context, compare choices or generate content within the room.

Group chats did not amount to a complete organizational communications system. The documented product lacked the broader operating layer associated with Slack, including organization-wide workspaces, extensive channel structures, centralized corporate administration and a mature ecosystem of workplace integrations.

The distinction matters because a useful shared AI thread and a company’s primary messaging infrastructure solve different coordination problems. Group chats tested whether an assistant could participate intelligently in a multi-person exchange; they did not establish ChatGPT as the system through which an organization would manage all internal communication.

What the short pilot demonstrated

The experiment showed that one AI context could incorporate preferences, corrections and source material from several people. ChatGPT could then summarize the combined exchange or respond to a shared request without each participant maintaining a separate copy of the discussion.

It also exposed coordination questions that are less prominent in a one-to-one chatbot: when the assistant should remain silent, whose request it is answering and whose usage allowance applies to a response. The product included mechanisms for those situations, but there is no public evidence tying any particular mechanism to the retirement decision.

The outcome is narrower than the early speculation. OpenAI released group chats globally, operated them as a pilot and then began withdrawing them. Existing rooms are becoming archives, while a successor capable of challenging established workplace messengers has not been announced.

ChatGPT Group Chats Went Global—Then OpenAI Closed New Rooms