According to Thurrott.com, OpenAI has begun testing a new group chat feature that lets multiple users collaborate in the same ChatGPT conversation. The feature is currently rolling out to mobile and web users in just four markets: Japan, New Zealand, South Korea, and Taiwan. Users can start group chats by tapping the people icon in any chat, creating separate conversations where personal ChatGPT memory is automatically disabled. Group members can invite up to 20 participants using share links, and participants must create profiles with names, usernames, and photos. The chatbot has been trained with new social behaviors to follow conversation flow and decide when to respond versus staying quiet based on context.
The geographic limitation problem
Here’s the thing – launching in only four markets feels incredibly restrictive. Japan, New Zealand, South Korea, and Taiwan? That’s a pretty specific testing ground that leaves out most of ChatGPT‘s user base. I get that companies need to test features gradually, but this feels like they’re being overly cautious. And let’s be honest – cultural differences in how people collaborate could mean the feature works great in these markets but completely bombs elsewhere.
Teaching AI social manners
The idea that ChatGPT has learned “new social behaviors” for group settings is both fascinating and slightly terrifying. Getting an AI to understand when to jump into a conversation versus when to stay quiet is basically teaching it human social cues. But how well does this actually work? We’ve all been in meetings where that one person just won’t stop talking – now imagine that person is an AI. The fact that you can mention “ChatGPT” to force a response suggests they’re aware it might not always get the timing right.
Safety and privacy questions
Automatic safeguards for users under 18? That’s smart, but it raises bigger questions about moderation in these group spaces. With up to 20 people potentially uploading files and images, who’s responsible when things go sideways? And disabling personal memory in group chats makes sense for privacy, but it also means the AI can’t build on previous interactions within that context. You’re basically getting a dumber version of ChatGPT in group settings.
Microsoft is already competing
Meanwhile, Microsoft just announced a similar groups feature for Copilot that supports up to 32 people – that’s 12 more than ChatGPT’s limit. But Copilot groups are only available in the US, which creates this weird geographic split where neither company has full coverage. It feels like both companies are rushing to match each other’s features without really thinking through the user experience. Basically, we’re getting an AI collaboration arms race where the actual collaboration part might be the least developed aspect.
Where this is heading
OpenAI says group chats are “just the beginning” of ChatGPT becoming a shared space, which suggests they have bigger plans. But I’m skeptical about whether people actually want to collaborate with AI in group settings. Most collaboration tools struggle with adoption as it is – adding an AI participant might just complicate things further. The real test will be whether these features actually improve productivity or just create another digital distraction.
