Blunt verdict first: Island didn’t double its valuation because enterprise browsers got more interesting, it doubled because AI agents turned every logged-in session into a liability nobody knows how to audit.
The numbers, for the record. On September 24, 2026, Island announced a $400 million Series F led by Evolution Equity Partners, pinning the company at a $6.4 billion valuation. That’s more than double where it sat in 2024. And the stated direction is a move beyond the enterprise browser into broader corporate systems, with AI agent security as the framing.
I review AI tools for a living, which means I spend a lot of time watching agents do things their operators didn’t quite intend. So let me tell you why this round makes uncomfortable sense.
The browser stopped being a browser
Island’s original pitch was simple enough to fit on a napkin: most enterprise work happens in a browser tab, so put the security controls inside the browser instead of bolting them on around it. Copy-paste rules, screenshot blocking, session visibility, the whole thing enforced where the work actually happens.
That pitch aged well, and then it aged strangely. The browser is now where AI agents live too. Agents click. Agents fill forms. Agents read the SaaS dashboard your finance team uses and then go write something somewhere else. The tab is no longer a window a human looks through. It’s an execution surface with a non-human operator holding the mouse.
Which reframes what Island is actually selling. It isn’t browser hardening anymore. It’s a control point for software that acts on a human’s behalf, with that human’s credentials, at a speed no human would produce.
Why investors are paying up
I’m usually the first person to call a valuation silly, and $6.4 billion for browser security would have sounded silly two years ago. Here’s the argument for why it isn’t.
- Agents inherit permissions nobody scoped for them. When an employee gets read access to a CRM, that’s one person with one attention span. When an agent gets the same access, it’s a process that can enumerate the whole thing in minutes.
- Traditional tooling reads agent behavior as a human. Identity systems, endpoint agents, and access logs were designed around people. An agent using a human’s session looks like a very productive employee.
- The session is the new perimeter. Not the network, not the device. The authenticated session, and whatever is driving it.
- Enterprises are deploying agents before they’ve solved any of this. That gap is the product opportunity, and it’s large.
Put those together and the funding stops looking like hype and starts looking like a bet on a category that got redefined underneath its incumbents.
Where I’d push back
Now the skepticism, because this is agnthq and I’m not here to applaud press releases.
Expansion is the hard part
Island says it’s moving from the enterprise browser into broader corporate systems. Every security company in history has said a version of this, and most of them discovered that the thing they were great at didn’t automatically extend. A browser is a contained environment you control. Broader corporate systems are a swamp of APIs, legacy services, and integrations owned by people who will not return your emails. Going wide means competing with vendors who already own those beachheads.
“AI agent security” is doing a lot of work as a phrase
I’ve reviewed enough products to know that the phrase can mean anything from real behavioral enforcement to a dashboard that counts agent sessions and charges you per seat. Island hasn’t published, at least not in what’s been reported, the specifics of what its agent controls actually do. Until someone independent tests enforcement against an agent genuinely trying to exceed its scope, the category label is a promise, not a capability.
Control points cut both ways
Putting one vendor between your workforce, your agents, and your SaaS stack solves a visibility problem and creates a dependency problem. That’s a legitimate trade, but it’s a trade. Buyers should price it as one.
What this means if you’re deploying agents
Ignore the valuation. It tells you about investor conviction, not about your risk. What the round genuinely signals is that serious money now believes agent oversight is a standalone budget line, not a feature of something you already own.
So the practical question for any team running agents in production is narrow and unglamorous: can you name every system your agents can currently reach, and would you notice within an hour if one of them started behaving strangely? If the answer is no, you have the exact problem a $400 million round was just raised to sell into.
Island has the position, the capital, and a category that grew toward it rather than away. It also has an expansion story that has broken better companies. I’ll judge the product when I can test it. For now, the interesting fact isn’t the $6.4 billion. It’s that the security industry has quietly accepted that the risky user in your org chart might not be a person.
🕒 Published: