Island launches a secure web browser built for AI agents at work

Island, a company that makes secure web browsers for businesses, announced on Oct. 29, 2026 a version built for AI agents. Agents are software programs that do tasks for you, such as filling out forms. The new product, called Enterprise Browser for Agents, lets them use web and company apps while the organization's security rules still apply.

What changed

Agents already work inside companies. Sales agents update customer records after calls. HR agents onboard new hires. Finance agents process invoices. To do that, they need a browser, just as people do. Island says companies often bolted controls onto consumer tools. That setup may create risks such as exposed passwords, data escaping from approved apps, and records open to deletion.

How it works

Island says it governs what happens in the browser, not the agent itself. When an agent calls the browser or reaches internal resources, the company's rules apply to where data moves, where credentials are used and what comes out. It runs on devices and in the cloud, so it covers local and remote agents. Island says employees and agents cannot turn the controls off.

The cost angle

AI models are billed by tokens, which are small pieces of text they read. Island says consumer browser extensions often rely on screenshots, and pushing a full web page through AI burns tokens at every step. Island owns its browser engine and says its page-reading tools cut token use and make actions more reliable. It also says it remembers repeated workflows, so the same steps may finish faster next time.

What is not known yet

This is a company announcement. The material gives no pricing, no availability details and no independent testing. The claims about security and savings come from Island. The Gartner Peer Insights mention reflects opinions of individual users, and Gartner says it is not a statement of fact. Island's CTO, Dan Amiga, predicts most business browsing will come from agents within a few years. That is a forecast.

What to do

• List which company accounts your AI agents can log into.
• Decide what each agent may see, change and delete.
• Ask any vendor whether you can review a record of what agents did.