The move. On Oct. 29, Island launched the Enterprise Browser for Agents. The idea is simple. AI agents need a browser to do work: read web pages, log into SaaS apps, fill in forms. Island wants every one of those sessions to run inside its browser, with the company’s policies applied. A month earlier, Island raised $400 million at a $6.4 billion valuation and started calling itself “the agentic control plane for enterprises.”
Why pivot now? Look at the human side of the market. PeerSpot tracks how much buyers research and talk about each product, which it calls mindshare. In its enterprise browser category, Island leads at 27.0%, but that is down from 35.0% a year earlier. Palo Alto’s Prisma Access Browser rose to 24.2% from 22.8%. That isn’t revenue data. It is a decent signal of who is on shortlists. And the shortlist is tightening around the platform.
The bundle problem. Palo Alto bought Talon, rebranded it Prisma Access Browser, and now Prisma Browser. It sells it next to its SASE stack, which is the cloud network security service that routes and inspects company traffic. One independent comparison puts the issue plainly: Prisma Browser is often priced as an add-on inside a bigger SASE deal, so it looks cheap. Island’s AWS Marketplace listing starts at $250,000 for a year, before per-user fees. When a CISO already pays Palo Alto for SASE, “good enough and already in the contract” is a hard pitch to beat.
And the extension problem. The other squeeze comes from below. Many security teams don’t want to swap the browser people already use. They want controls inside Chrome and Edge. Push Security, which sells a browser extension, ranks its own approach first in its 2026 roundup. That ranking is a vendor’s view, so discount it. But the direction is real. Island itself now sells an extension for Chrome, Edge, Firefox and Safari. When the category leader ships the thing that competes with its own core product, that tells you something.
Why agents are the smart bet. Humans complain when you change their browser. Agents don’t. Nobody has to retrain an AI agent or fight about bookmarks. Meanwhile, the risk is new and loud. An agent with a logged-in browser session can read anything a user can, click anything a user can, and do it thousands of times faster. Security teams have almost no record of what agents did inside SaaS apps. A controlled browser is one of the few places you can see and stop it step by step.
The catch. Every major platform is chasing the same agent problem. Identity vendors want to give agents their own logins. SASE vendors want to inspect agent traffic. Data security vendors want to limit what agents can reach. Island’s agent browser wins only if the browser turns out to be where agent actions are easiest to see and control. That is plausible for web and SaaS work. It is less clear for agents that call APIs directly and never open a page.
The valuation read. $6.4 billion is a platform price. Island has not disclosed revenue, so you can’t check the multiple. What you can check is the story. A replacement browser for contractors and VDI is a solid niche. A control point for every AI agent in the enterprise is a much bigger market. Investors are paying for the second one. Island now has to prove it before Palo Alto, Google and Microsoft bundle the same feature into contracts customers already sign.
For buyers. Don’t pick an enterprise browser for the human use case alone in 2026. Ask every vendor what happens when an agent, not a person, opens the session. Who logs it? Who can stop it mid-task? Can the vendor show you, not tell you?
Where do you think agent control ends up: in the browser, in identity, or in the network?