Island unveils secure browser platform for enterprise AI

Island has launched a set of enterprise AI products that put security and governance controls inside the browser and desktop environment where employees use AI tools.

The products are AI Protect, AI Browser, AI Automation and AI Publish. The suite is designed to address data leakage risks, policy violations and the lack of auditability many organisations face as staff adopt consumer and enterprise AI services.

Many companies have struggled to balance employee demand for AI tools with controls over sensitive data. AI use can fall outside established security models, particularly when staff use personal AI accounts or paste corporate information into chat interfaces.

Island cited analyst research on the nature of the risk. Gartner has warned that through 2026, 80% of unauthorised AI transactions will stem from internal policy violations such as information oversharing, rather than malicious external attacks.

Controls at the point of use

Island's approach focuses on what it calls governance at the point of use. Its platform can distinguish between work and personal AI tenants, set boundaries on what data can be shared, and apply policies before information reaches an AI provider.

AI Protect applies these controls across a range of AI entry points, including web-based tools and desktop applications. It also covers browser extensions and connectors that can move data between services.

The product captures audit logs of prompts and responses, along with activity from AI agents. It also applies browser-level protections against prompt injection and risky extensions. Prompt injection is a technique that attempts to manipulate a model or agent into exposing data or taking unintended actions.

"Today's AI delivers power without guardrails. Without visibility and governance, advantage quickly turns into exposure," said Dan Amiga, CTO and co-founder of Island.

"Island makes AI enterprise-ready, securing it in the browser, in AI apps, and across the desktop," Amiga said.

Browser integration

AI Browser embeds AI functions inside Island Enterprise Browser, Island's managed workspace for employees with centralised policy controls for IT and security teams.

Organisations can integrate more than one AI provider into a single chat experience. The setup supports frontier models and corporate data sources, including an option to connect AI to AI-enabled corporate data using retrieval-augmented generation (RAG).

AI Browser also adds enterprise context to AI interactions, which can include user role and content from business applications, along with approved company knowledge sources. Central policy and data protection controls still apply when employees use the integrated chat.

AI Browser also offers visibility into AI usage and costs. Organisations can track agent performance and measure time saved.

Agent oversight

AI Automation focuses on autonomous and semi-autonomous agents that act across enterprise applications. Teams can build and run on-demand agents with defined permissions and full audit trails.

Administration controls include limits on application access and the actions an agent is permitted to take. The product also supports human approvals for certain steps, a common requirement for tasks involving finance systems, customer records or administrative privileges.

Agents run in what Island describes as a hardened browser environment designed to reduce exposure to prompt injection and misuse.

Publishing workflow

AI Publish targets organisations building internal AI applications. It provides a way to deploy those applications under the same identity and security policies used elsewhere in Island's platform.

Published applications inherit identity enforcement, data protection, monitoring and policy controls. The product is aimed at teams that want internal distribution without setting up separate governance frameworks for each new app.

Platform approach

All four products run on the Island Enterprise Platform, which Island describes as a single environment for managing enterprise AI, with one control plane for policy, enforcement, visibility and auditing.

Island also outlined three components that deliver controls across common work environments: Island Enterprise Browser for managed browsing; Island Extension for controls across other browsers, including AI browsers; and Island Desktop for extending policies to desktop applications, including standalone AI tools.

Dentsu is among the customers using the platform, according to Island. The advertising and marketing group's security leadership highlighted the need for oversight of real-world AI use across teams.

"Island gives us the visibility into how our people use AI, the data to allow us to make key decisions, controls to scale it responsibly, and the ability to connect every team to the right AI, for the right purpose, at the moment they need it," said Paul Hennessy, global CISO at Dentsu.

"With great power comes great vulnerability. As AI evolves to interact with workers all day long, governance gaps will expand unless CIOs and CISOs can convert consumer AI into enterprise-grade AI," Amiga said.