Brave Isolated AI Browsing Profile

Brave is trying to make the browser itself the safe sandbox for AI agents, instead of bolting an agent onto a normal tab and hoping permission prompts are enough. Its AI browsing setup starts in a separate profile, blocks access to regular tabs, internal browser pages, non HTTPS sites, and unsafe domains, which turns the browser into both the agent interface and the security boundary.1 2 3 4

• This matters because agentic browsing means the model is not just answering questions, it is clicking through pages, reading forms, and taking actions. Brave is framing that as a browser feature, inside Leo, rather than a standalone operator product, which keeps the workflow inside Brave’s existing privacy stack.1 3 5
• The isolated profile design is the concrete differentiator. If the agent cannot see the user’s normal logged in tabs, cookies, browser settings, or local browsing state, Brave reduces the biggest failure mode of browser agents, which is accidental access to sensitive session data while the model is exploring the web.1 2 4
• This also fits Brave’s broader business logic. With more than 100 million monthly active users and browser revenue at a $100M annualized run rate in April 2025, adding useful AI actions inside the browser gives Brave a way to deepen engagement without abandoning its privacy position, unlike ad driven browsers that monetize more tracking by default.6 7

The next step is a browser that does more of the work of using the web, while keeping the agent fenced into tightly controlled environments. If Brave can make agentic tasks feel useful without exposing sessions, tabs, and identity data, it has a credible path to turn privacy from a blocking feature into the foundation for an AI native browser.1 2 3 5