OpenAI today introduced a new "Workspace Agents" feature on the ChatGPT platform, available in research preview for enterprise and team users. This functionality is seen as a significant upgrade to the existing GPT framework, positioned as an extension rather than a replacement for custom GPTs. OpenAI clarified that existing GPTs services will remain operational, with future tools planned to support seamless migration of existing GPT configurations to Workspace Agents. According to technical documentation, Workspace Agents enable automated task processing across tools by integrating Operator's interaction capabilities with Deep Research's in-depth analytical abilities. Users can trigger agents via natural language instructions to autonomously complete complex operations such as web browsing, data scraping, code execution, and file generation, with direct support for editable content output in productivity tools like Excel and PowerPoint. In terms of security mechanisms, the agents employ a multi-layered protection system, including operation confirmation mechanisms, sensitive command identification, real-time trajectory monitoring, and manual user takeover functions. OpenAI emphasized that this feature remains in research preview and advised against its use in high-risk scenarios or for handling sensitive information. Enterprise users can apply to participate in the preview program through OpenAI's official channels, with plans to gradually expand functional coverage and user accessibility in the future.