ChatGPT Enables Group Collaboration

What happened

OpenAI introduced a pilot group chat feature within ChatGPT, enabling up to 20 logged-in users across all subscription plans to collaborate directly with its AI via a shareable link. Available in Japan, New Zealand, South Korea, and Taiwan, this functionality is powered by the GPT-5.1 Auto model, which dynamically selects the AI based on prompt and user plan. Participants can utilise image generation, file uploads, and web search, with rate limits applying solely to ChatGPT's responses. Group members gain administrative controls for renaming, adding/removing participants, and setting custom AI instructions.

Why it matters

The introduction of shareable links and user-managed group settings creates a control gap for platform operators and IT security, increasing exposure to unauthorised access and content management challenges. Unconstrained user-to-user communication, due to rate limits applying solely to AI responses, reduces visibility for compliance teams regarding internal information flow and potential policy violations. The automated GPT-5.1 Auto model selection also introduces a visibility gap for audit and governance teams concerning specific AI model behaviour and data handling within collaborative sessions.