Most organizations are stuck: employees want to use AI, but IT teams lack the controls to allow it safely. Island's launch event introduced four new capabilities designed to close that gap.

AI Protect gives security teams full visibility into AI usage across the enterprise and the controls to govern it. AI Browser brings a governed, customizable AI experience directly into the browser — with data protections enforced in real time. AI Automation lets users run agentic workflows across business systems, with human oversight built in. And AI Publish lets anyone deploy a custom AI app to the entire organization in minutes, without sacrificing security or compliance.

The throughline: enterprises no longer have to choose between enabling AI and protecting their data.

Read on for a transcript of the AI launch event.

Steve Tchejeyan AI is coming to the workforce. It's already here today, and there's going to be more and more tools for CIOs and CISOs to say yes.

Roy Rashti There is a big tension between the business and the IT department today because IT teams don't have the necessary controls to safely enable that in the enterprise space.

Bill Van Kuren Island will help the CIO understand where AI is being used throughout the enterprise, all the way down to the desktop of the individual user.

Revital Rubin You have one policy that covers your users, but also covers all the AI tools, the extensions, the chat providers, the agentic processes, and the agents they are using.

Mike Fey Let's turn our attention to AI. We saw AI's value to the end user years ago. We built in solutions to leverage and deliver that to the user. The role we now play is AI enablement.

What does that mean? When you think about delivering AI to the end user, we have to protect them, our company, our data, our customers — all that we value. Island provides a unique and efficient approach to protecting them, but we don't just stop there.

We take that advantageous position of being on every endpoint, seeing everything that occurs, and we elevate that yet further. We have built automation into the platform. We connect you to your AI properties that you've invested in and manifest those on the endpoint to drive additional value.

Beyond automation is Publish. What we do with Publish is give you the ability to think about the endpoint like you thought about Kubernetes and containers in the cloud. What if we write this code quickly? What if we deploy this AI opportunity to the endpoint? Do I have to build every feature to make it enterprise-grade? No, you don't.

Think of Island as the safe passage. If you put it there, we will wrap it with identity and security, we'll connect it into the infrastructure so automation is easier, we'll give you visibility, compliance, everything you need — so that your developers stay focused on adding value to the process, not supporting all the other requirements of an enterprise application.

02:20
Dan Amiga If you think about OpenAI, Anthropic, Gemini, hosted models and the like — the tech is the model itself. When you need to massively deploy and roll it out in an organization, you need security, data protection, insights and workflows, and policy. That's what the platform provides.

Island lets the organization choose. Many have already chosen a supported LLM vendor for their workflows. One might say, "I'm going to go with Anthropic for financial workflows, and Gemini for everything else." We give them the enterprise tools to roll it out.

Let's start with AI Protect. AI Protect delivers visibility and control across all AI tools. Customers need a risk reputation engine on extensions to understand what AI extensions are being used. They need visibility into the desktop and into the network.

Roy Rashti In the hierarchy of needs for an organization that wants to enable AI, this is the basics. They have to know how many users are using AI, what AI applications are being used, how many prompts are sent, and how users are signing into those applications — whether with a personal or enterprise account.

Island can surface very interesting insights: sensitive information being transferred to a non-corporate account, large volumes of data being uploaded to an AI application, new AI applications appearing in the organization. These are unique capabilities because of how much data and visibility we have in the platform.

We also have visibility into how files move within the organization. We can show clipboard actions — where clipboard data was moved from and to with respect to AI applications. And because we are also the network, we can show insights into MCPs and tools. If a tool moved data between Slack and ChatGPT, we have visibility into that because we control the presentation layer — we're the browser, we're at the desktop.

AI extensions have been a growing risk for organizations in 2025, and it's likely to be an even bigger risk in 2026. Island is well positioned here. We can show customers exactly what extensions are being installed in the organization, including risk scores and permissions, to help them evaluate possible risks. We can help them set policies to limit the extensions being deployed.

Beyond visibility, we provide control. Our 360 DLP policy — spanning desktop, network, browser, and extensions — can be applied to AI applications, along with very specific controls for Gemini, ChatGPT, Copilot, Claude, and more. It's a library that keeps growing.

Dan Amiga One product. We can see everything users are doing in the enterprise browser. We get full visibility by rolling out the extension on other browsers. We get desktop visibility. Where other vendors rely on blocking, we allow organizations to understand what is being used, what the risk is, what data is moving between existing apps and AI apps — and then create a governance policy around that. That is AI Protect.

Now, AI Publish. This enables organizations to take consumer-generated apps and wrap them in an enterprise envelope, then deliver them at mass scale. HR teams want to create apps. Sales teams want to create apps. AI Publish takes that and pushes it into the enterprise. Let's see a demo.

08:28
Roy Rashti The thing about AI Publish is that the distance between going to Lovable to create an application — which takes about five to six minutes — and putting that application in the hands of end users is identity, data protection, monitoring, and visibility. With Island, we shrink that distance to just a couple of minutes in the management console.

Here's how it works. In the management console, you create new content and choose an AI widget, application, or tool depending on what you're trying to achieve.

Let's say I want to build something that addresses gaps in Salesforce — things Salesforce doesn't do today that we need it to do. If we asked Salesforce to build that, it might take months, years, or never happen. So we do it ourselves. We take the code from Lovable, choose an icon, and add it to the management console. I can select the exact audiences I want to apply it to. That's it — two minutes of work after the seven minutes it took Lovable to generate the code.

Once the policy is pushed, users immediately see the new widgets on their homepage. In this example, we've built a calendar widget and an email widget. But it can be anything. Our product team built a world clock widget because coordinating across time zones is always a challenge.

Dan Amiga When you think about the customer experience on the homepage, they are no longer bounded by what enterprise communication used to offer — a survey, a poll, whatever. You can generate any app. And what Roy is showing is real-world integration. These are his actual emails and meetings on the homepage.

Roy Rashti Here's the Salesforce helper. On the left side is the internal legacy application — Salesforce, a SaaS app we can't fully control. On the right side is our own application that does exactly what we need Salesforce to do, but doesn't today. That's how we bridge the gap.

In this example, we use AI insights to connect to internal systems, fetch data, surface what we need, and enable quick actions like outreach or data export — all placed right next to the existing business application.

Dan Amiga This is the first time in software where you can go to existing apps — where companies have invested so much in customization and change management — and build an augmentation layer super fast. The AI app on the right understands the page. How long did it take to build? A few minutes. And we've augmented Salesforce with the organization's business logic in a few minutes, with everything secured, audited, and policy-applied.

Roy Rashti And because it's so easy, it's feasible to do this across many business applications without a development team. We built another application on Gmail in just a few minutes. On the left is the email. On the right is a tool that tells you who you're emailing with, what account they're associated with, talking points, and a thread summary. A real lifesaver, built in minutes.

14:50
Dan Amiga Now, the AI Enterprise Browser. This isn't a new product — organizations can choose whether to enable the AI capabilities. Gartner recently stated that cybersecurity teams must block AI browsers for now, because for these browsers to work, they send as much data as possible to the LLM vendor for better context and results. That's the problem.

What we did is take the main pillars of an AI browser and make them enterprise-ready. If you roll out OpenAI's browser, you're married to OpenAI's LLM. If you use Chrome, you're married to Gemini. But what if Claude is a better model for your engineers? This isn't like the early search engine days where there's one primary truth. Organizations need flexibility.

In 2025 alone, there were periods where Grok excelled at sentiment analysis due to its Twitter integration, Gemini outperformed others on certain tasks, and Anthropic became the de facto choice for software engineers. Organizations need to bring their own model. They need data protection controls — what can users type in, what can be shared with the cloud. They need enterprise context — who the user is, their job function, what business applications they connect to.

So we implemented enterprise chat, an enterprise agent within the browser, enterprise context, and many more capabilities. And we made it beautiful. We worked hard on the edges, the latency, the animations — and we're improving it every day.

Roy Rashti Here are the new AI browser features. We have a new Omnibar at the top and an embedded chat — which can be any chat: our enterprise chat, Copilot, Gemini, Grok, Claude, ChatGPT, you name it.

What makes this unique is the ability to customize and tailor the experience. Think about doctors — AI could make them significantly more efficient and faster. But the way AI browsers work today makes that nearly impossible to implement safely. What I'm about to show you is unique for a few reasons.

First, customization. The browser is branded, and the entire experience is branded. We can suggest smart actions based on the context of the website and the identity and role of the end user — because a doctor and a nurse may need two different things when viewing the same medical record.

When I ask the system to analyze the vitals on this page, Copilot attempts to access the page content. But Island acts as an AI web application firewall and blocks the request because it violates the organization's DLP and data boundary policies.

Here's what makes it really interesting: we can distinguish between situations where sensitive information is present and where it isn't. We can mask the data on the page — the name, ID number, and SSN are no longer present in the DOM. When I ask again, Copilot tries to access the page content, we evaluate it against our data policy, and because there's no sensitive information, it's able to proceed.

20:23
Dan Amiga Now let's walk through AI Automation — how to create workflows that span different websites or services. AI can take a customer process from 10 minutes to 10 seconds.

All AI vendors offer some form of automation, but there's a meaningful difference between suggesting a hotel booking and creating and executing an enterprise transaction. The accuracy requirements are different. The policy requirements are different.

There are a few methods we use: analyzing the page structure, reading the page content as an image the way eyes would, and identifying more deterministic approaches like API calls or MCP servers.

Roy Rashti Let me show you how easy it is to create a workflow. You have a library of built-in skills. To create a workflow, you simply prompt it. For example: get all of my leads from Salesforce, then send a cold outreach email about Island Enterprise.

The system thinks through the best course of action using the available skills. If I want to refine it — say, only contact leads I haven't reached before — I just add that instruction and save.

When I run it, the agent evaluates the organizational policy, confirms what the user is entitled to do, contacts Salesforce via API, fetches leads, uses AI to determine which ones haven't been contacted, and then keeps me in the loop for any sensitive action. I select the leads I want to process, review the email before it's sent, approve it, and it goes. All while I continue working in other tabs.

We've also built in ROI tracking — one of the biggest challenges in AI enablement. We can show a detailed view of which users are running which workflows, success rates, time to complete, and most importantly, how much time was saved.

Dan Amiga What AI did is make it easy to create regular workflows. What we did is leverage that to create more deterministic workflows, tie them into enterprise services, and deliver them through the Island browser. Real examples include automating media publishing in Amazon, logging into Facebook to learn the latest issues, updating account permissions, updating PO numbers, and resolving incidents.

Roy Rashti This is real. It's happening. It's easy to deploy. And the product is getting better every day.

26:39
Mike Fey We have taken the end user experience very seriously. As you work with us, you'll find that rare opportunity to have a better end user experience while reducing your cost structure — and doing so securely.

Kelly Burns I just had someone come to me and say, "This Island thing changed my life." To me, it's a platform, an ecosystem — a space where you can do so much more.

Dimitri Alperovitch It's literally the first product I've ever seen with security use cases where the users actually love it. It's empowering users to be more productive with AI integrations and all the other capabilities that Island delivers.

Dan Amiga This is the future of enterprise work, and it just arrived early. Welcome to the Island Enterprise Platform.