Island introduced its Enterprise Browser for Agents on October 29, describing a way for organizations to apply browser security policies when AI software works across business applications and websites.
The company says agents connect through a Model Context Protocol interface and operate with a user’s enterprise identity and permissions. Administrators can apply policies to people, their agents or both.
The controls can distinguish between reading information and changing it. Island says organizations can limit which pages and browser tools an agent may use, including allowing access to sensitive systems without permission to submit, modify or delete records.
The announcement also describes controls over credentials and data movement, plus audit records identifying whether a person or an agent performed an action. Organizations can optionally record sessions for playback.
Island says the browser supports agents running on devices and in the cloud. Its stated scope is governing browser activity, rather than controlling every operation performed by an AI model or agent.
Sources: Island — October 29 Product Announcement, Distributed by GlobeNewswire: https://portal.sina.com.hk/finance/finance-globenewswire/globenewswire/2026/10/29/1971450/island-brings-the-enterprise-browser-to-ai-agents/