Abstract
Incognito states or modes can be managed automatically during the execution of agentic browser flows to address data management concerns during sensitive tasks. When a user requests an AI agent to perform a task, a sensitivity level of the task can be determined based on stored confidentiality preferences. These confidentiality preferences can be derived, with user permission, from on-device analysis of historical behavior by the user or from global data that includes, with user permission, numerical features describing the preferences of other users. If the task, the execution plan of the task, or specific Universal Resource Locators (URLs) associated with the task are determined to meet a sensitivity threshold, the AI agent executes the task within an incognito browsing session, which prevents local history logging and third-party tracking. The user can optionally be prompted for confirmation before initiating the incognito session, ensuring transparency and alignment of automated tasks with user intentions.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Azose, Benjamin Albert and Azose, Jonathan Jerome, "Determining Whether to Enter Incognito Mode Based on AI Agent Request", Technical Disclosure Commons, (May 11, 2026)
https://www.tdcommons.org/dpubs_series/10062