ChatGPT launches pilot group chats across Japan, New Zealand, South Korea, and Taiwan (techcrunch.com)

OpenAI has begun piloting a group chat feature in ChatGPT across Japan, New Zealand, South Korea and Taiwan, available to Free, Plus and Team users on web and mobile. The invite-only groups (1–20 people) let multiple humans converse alongside ChatGPT—powered by GPT-5.1 Auto—with search, image generation, file uploads and dictation available in-group. Starting a group is as simple as tapping a people icon or sharing a link; adding someone to an existing conversation spawns a new group so originals remain unchanged. Privacy controls are emphasized: private chats and personal memory remain private, groups are invitation-only, members can leave at any time, most participants can remove others (the group creator can only leave voluntarily), and extra safeguards/parental controls filter content for users under 18. Importantly, ChatGPT’s hourly response limits only apply when the AI responds, not for human-to-human messages, and users can tag the assistant to prompt it or have it react with emojis and personalized images. For the AI/ML community this marks a concrete move from single-user assistants toward collaborative, multi-agent experiences and social features. Technically, it surfaces engineering challenges and opportunities around multi-party context management, turn-taking, identity and access controls, moderation, and data governance—especially since group interactions may produce new shared datasets and require refined safety pipelines. The pilot’s feedback-driven rollout will be watched closely for how model behavior, rate limits, and privacy guarantees scale in collaborative settings and for signals about future social products (e.g., Sora 2) that blend generative AI with social UX.