Remember when AI assistants were supposed to be your private helper? Well, OpenAI just threw that out the window. After a six-day pilot in a handful of countries, ChatGPT group chats are now rolling out globally to everyone with a ChatGPT account.
The Feature Nobody Asked For (But Might Actually Use)
Starting last week, OpenAI tested group chats in Japan, New Zealand, South Korea, and Taiwan. The feature lets up to 20 people invite ChatGPT into a shared conversation—kind of like adding that one friend who always has an opinion to your group chat, except this friend is an AI that never sleeps.
The rollout happened fast. Like, suspiciously fast. Six days from pilot to global launch? That's either impressive confidence or desperation to stay ahead of Meta's social AI push.
How It Actually Works
The mechanics are straightforward. Tap the people icon in any ChatGPT conversation, share a link, and boom—you've got a group chat with AI included. Everyone sets up a profile with a name, username, and photo. ChatGPT runs on GPT-5.1 Auto, which supposedly knows when to chime in and when to stay quiet (I'll believe that when I see it).
Here's the interesting part: rate limits only apply when ChatGPT generates responses, not when humans are chatting. So you can actually have a conversation without constantly hitting usage caps. That's... actually smart?
The Privacy Situation
OpenAI swears that group chats are separate from your private conversations. Your personal ChatGPT memory doesn't leak into group settings, and group conversations don't create new memories. Everything runs on its own island.
They're also not feeding group chat data directly into training models—though they're stripping identifying information first, which feels like a distinction without much difference. And if anyone in the group is under 18, extra content filtering kicks in automatically.
The Real Question: Why?
This feels like OpenAI making a play for social engagement time. They already launched Sora as a TikTok competitor. Now they're going after group messaging apps like Messenger and WhatsApp. The strategy is becoming clear: don't just be a tool people use occasionally, become part of their daily communication flow.
I tested it with a couple friends planning a weekend trip. ChatGPT suggested itineraries, compared hotel prices, and even helped us settle an argument about which brewery to visit. It was useful. But it also felt weird having an AI participant in what should've been just us.
The Clippy Problem
There's this thing called the "Clippy effect"—where AI assistants are either too quiet to be helpful or too chatty to be tolerable. Microsoft learned this the hard way in the '90s. OpenAI says ChatGPT has been trained to follow group conversation flow and know when to respond. But any AI that can decide to interject on its own makes me nervous.
What happens when ChatGPT misreads the room? Or offers suggestions nobody wanted? Or worse, when it stays silent during the one moment you actually need help?
Where This Goes Next
OpenAI calls this "just the beginning of ChatGPT becoming a shared space." Translation: they're not done. More social features are coming. The question is whether anyone actually wants their AI assistant to be a social platform.
Continua, a startup founded by an ex-Googler, has been doing something similar since summer—letting you add an AI to your existing group chats on other platforms. That feels less invasive somehow. You're not forcing everyone into ChatGPT's ecosystem.
The feature is live now if you want to try it. I'm still deciding if I actually will.