Chat mode allows users to ask questions about their open browser tabs and receive answers based on the information displayed across multiple webpages. For instance, users comparing prices across four retailer websites can ask ChatGPT to evaluate product specifications, delivery windows and return policies, with the tool synthesising the information into a direct response.
Work mode introduces browser interaction capabilities. According to reports citing OpenAI’s announcement, the feature can fill out forms using information extracted from open webpages or files stored on a computer, send messages, download files and organise open tabs into topic-based groups.
Reporting by Analytics Insight confirms that users can review completed forms before submission. However, available information does not establish that every browser action requires manual approval, making it important for users to understand the distinction between reviewing a completed form and authorising other actions.
OpenAI has also introduced a shortcut that reportedly allows users to invoke browser tasks directly from the ChatGPT website or app. By typing “@Chrome”, users can reportedly access browser-related tasks without opening the extension’s side panel.
What Users Need to Know Before Granting Access
The extension requests permission before accessing browser history, visiting new websites or handling files. According to reports, users can grant permission once, on a per-site basis, across all sites or decline access entirely.
Users can also restrict the extension to specific tabs, limiting the information available to ChatGPT during a task. Reports citing OpenAI’s announcement identify Gmail, LinkedIn and Salesforce as examples of services whose signed-in sessions the tool can interact with when broader access is granted.
These permissions carry practical implications because Work mode can send messages and populate forms. If users grant broad access, the tool may be able to interact with active sessions on services they use regularly, making the scope of each permission an important consideration.
Browser compatibility is another consideration. The extension reportedly supports Chrome, Edge, Brave and Opera, with Vivaldi also mentioned in some reports. However, sources are not fully consistent about Vivaldi’s inclusion. Firefox and Safari were not included in the initial rollout, according to available reporting.
The availability of browser-control features across ChatGPT subscription plans also remains unclear. Reports differ on whether access to browser control requires a paid plan, leaving uncertainty about which users can access the full range of Work-mode capabilities.
Setting up the extension involves using the ChatGPT desktop app. Users must open Settings, select Computer Use, install the extension for their preferred browser and then select that browser within a ChatGPT Work or Codex conversation.
The update marks a step in OpenAI’s effort to expand ChatGPT from an assistant that provides information into a tool that can carry out tasks in a browser. Rather than simply answering questions or retrieving information, Work mode is designed to navigate websites and perform actions within the limits of the access users grant.
Its effectiveness in completing multi-step tasks and the practical reliability of its permission controls remain open questions. Whether the extension becomes a useful productivity tool or introduces additional risks will depend in part on how accurately it performs tasks and how carefully users manage access to their browser sessions, files and online services.
About The Author
Comment List