Open-tab content is the information currently loaded in a browser’s tabs, such as text, documents, or data displayed in a web app. A browser assistant may use this content as context to answer questions or perform tasks, potentially without the user copying it into a chat.
This matters because tabs can contain confidential information or provide access to logged-in services. If an assistant can read page content, that data may enter its workflow; if it can also control tabs, it may take actions using the browser session’s existing privileges. Defenses include reviewing what permissions an assistant has, separating sensitive browsing from general use where practical, and checking actions before they are completed.