This isn’t just another feature. It’s a strategic move to turn ChatGPT from a solitary assistant into a full-blown collaborative partner, and it’s powered by a brand-new, “socially aware” AI.

If you’ve ever found yourself in the digital black hole of a group chat—endlessly debating dinner plans, trying to coordinate a vacation, or just losing a work project to a sea of memes—OpenAI is building a new life raft. The company has officially started pilot testing group chats directly within ChatGPT, fundamentally changing the AI from a one-on-one tool to a participant in our collective chaos.

For now, this new world is only available to users in Japan, New Zealand, South Korea, and Taiwan. But the implications are global.

The concept is brilliantly simple and instantly familiar. Like any group chat on your phone, you can create a conversation with friends, family, or colleagues. The key difference? ChatGPT is right there in the chat with you, acting as an active, intelligent participant.

The use cases write themselves.

• Planning a night out? Instead of 10 people shouting, “I don’t care, you pick,” you can have the group dump their preferences and ask ChatGPT to find a restaurant everyone will actually enjoy.
• Organizing a vacation? The AI can build a full itinerary, compare destinations, and create a shared packing list as your friends argue over “beach” vs. “mountains.”
• Collaborating at work? This is where it gets really powerful. You and your colleagues can upload articles, notes, and data, then ask ChatGPT to outline a report, summarize the key findings, or draft an email based on the entire group’s input.

How it works (and the new brain behind it)

Getting started is simple. On any new or existing conversation, you tap a new “people icon” at the top right of the screen.

If you start from an existing chat, don’t worry—your original one-on-one conversation is safe. ChatGPT simply duplicates that chat history into a new group conversation, giving everyone new immediate context. From there, you can add people by sharing an invite link.

Here are the nuts and bolts:

• Group size: You can have from one to 20 people in a single chat.
• Profiles: To avoid confusion, everyone who joins has to set up a quick profile with their name, username, and a photo.
• Admin controls: The chat creator has ultimate control, but other participants have power, too. Anyone who has the link can invite more people, and participants can mute or remove other members from the chat (except for the creator, of course).

But the coolest part isn’t the interface; it’s the AI powering it. Group chat responses are run by GPT-5.1 Auto. This isn’t just a new number—it’s an entirely new system. “Auto” means the AI intelligently and automatically switches between different models based on your prompt. For simple, fast questions, it might use “GPT-5.1 Instant.” For a complex request, it will switch to “GPT-5.1 Thinking” to apply deeper reasoning before answering.

More importantly, OpenAI says it has specifically trained the chatbot to understand the flow and social cues of group conversations. It’s designed to know when to stay quiet and let the humans talk, and when to jump in with a helpful answer. If you need it to pay attention, you can just summon it by mentioning “@ChatGPT” (just like you’d @-mention a person in Slack or Discord).

Let’s be honest, the first question everyone has is: “Can my friends see my private chat history?”

OpenAI is way ahead of this. The company is being extremely clear about the privacy guardrails. Group chats are entirely separate from your personal conversations. Your personal “Memory” in ChatGPT is not shared with the group, and nothing that happens in the group chat is used to create or update your personal memory. The AI in the group only has access to the messages and files shared within that specific group.

This privacy-first design isn’t an accident. It’s a critical strategic move. In a world where users are (rightfully) paranoid about data privacy, this walled-garden approach is the only way such a feature could work.

This launch also comes at a time when OpenAI is publicly and aggressively defending its user privacy in a legal battle with The New York Times, stating that it will not turn over private user conversations. Launching a new social feature with such strong, clear privacy boundaries is a powerful statement.

The second reason this matters is the competitive landscape. This feature is a direct shot across the bow at… well, everyone. It’s competing with the AI in Meta‘s WhatsApp, it’s competing with collaboration tools like Slack, and it’s part of a larger push to make ChatGPT the “everything app” for an AI-powered world. It’s a shift from AI as a tool to AI as a teammate.

For now, it’s just a test. The company says it will be tweaking the feature based on feedback from these first early users. But if the pilot is successful, get ready for your group chats to get a whole lot smarter.