ExSitu

Cloudflare Launches Dedicated Browser for AI Agents: Kitesurf

According to a TechCrunch report, Cloudflare has introduced a cloud-based browser called Kitesurf, designed for AI agents rather than human users. Developers building browser-based automations may find it a relevant tool — but organizations need to evaluate it before putting it into actual use.

2 min read

Cloudflare Launches Dedicated Browser for AI Agents: Kitesurf

According to a TechCrunch report, Cloudflare has launched a new tool called Kitesurf — a cloud-hosted browser designed from the ground up for use by AI agents, not humans. According to the report, the company claims the browser consumes fewer computing resources than standard Chromium browsers when performing common automation tasks, which is meant to help developers build browser-based AI agents more efficiently.

The tool is aimed mainly at developers and engineering teams building AI products that include an automated web-browsing component — for example, agents that perform tasks such as gathering information, filling out forms, or interacting with websites on a user's behalf. For mid-sized Israeli organizations, the direct relevance at this stage is limited mostly to internal development teams or technology vendors that incorporate AI agents into their products, rather than day-to-day office use.

The report does not specify what happens to data entered into the browser or collected by AI agents operating through it, and makes no explicit reference to a privacy policy or data storage. In the absence of such information, this tool should not be treated as a safe default for handling sensitive or organizational data, and this should be verified directly with Cloudflare before any actual use.

🔴 Key risk: any new AI tool that enters an organization — including a browser designed for agents — constitutes Shadow AI as long as no clear usage rules have been defined: who approves its activation, which systems and data it can access, and how the activity it performs on the organization's behalf is documented. A developer who installs such a tool to build automation, without informing IT management, creates an uncontrolled access point to internal data and systems. Presenting a tool of this kind without reasoned caveats is effectively a recommendation to put it into use — so any adoption decision should go through an orderly review process, rather than being based on initial enthusiasm for a new capability.

At this stage, this is a relatively new tool, and the report does not provide full detail on security mechanisms, access controls, or support for local regulatory requirements. Organizations considering using it, whether directly or through a vendor that incorporates it, should review the matter with Cloudflare individually before making a decision.

What This Means for Your Organization

  • A new AI browser tool does not automatically become safe for organizational use — independent review of its data policy is required
  • AI agents that perform automated browsing on an organization's behalf require clearly defined permissions and access set in advance, not after the fact
  • Developers who adopt such tools on their own, without IT's knowledge, create a classic Shadow AI risk

Sources