Cloudflare launched a browser designed for AI agents rather than for human use, reported in the first week of August.
The idea answers a concrete problem seen across several incidents in the period. When an agent uses a person's browser, it inherits open sessions, authentication cookies and access to everything that person was logged into, which turns any redirection into access to a real account.
A separate environment allows giving the agent its own credentials, defined scope and audit logging, the same direction taken by agent identity products announced weeks earlier.
What the separation doesn't resolve is the root of the problem. The agent still reads content from pages anyone can write, and still can't reliably distinguish data from instruction. Research from the same period was explicit that there is no perfect fix.
The useful reading is that the market began treating automated browsing as its own category, with its own requirements, rather than adapting a tool built for people.