Need smarter insights in your inbox? Sign up for our weekly newsletter to get only the things that matter to enterprise AI, data and security leaders. Subscribe now
An exclusive interview with Sam Evans of Clearwater Analytics’ CISO BentureBeat reveals why in many ways the enterprise browser is quickly becoming the forefront defense against Shadow AI.
Evans faced a major challenge in October 2023. Standing in front of Clearwater Analytics’ board, they had to confront concerns that employees could inadvertently publish data that could potentially infringe on $8.8 trillion in assets under their control.
“The worst is one of those employees who get customer data and put it in an AI engine that doesn’t manage,” Evans told VentureBeat. “Employees either don’t know what’s different or are trying to solve a customer’s problem. That data helps train the model.”
Here’s the conversation with Evans, edited for length and clarity
VentureBeat: What do you think AI today is shaping cybersecurity?
Evans: The attack is very refined. Given that from a bad actor’s perspective, the phishing emails and attempts we receive have become much more complicated. However, AI also has the ability to respond.
I like to explain it to our board as the ultimate cat and mouse game. It helps the other side of security practitioners use AI to facilitate our response, as it encourages the time it takes for bad actors to use AI to advance phishing or an exploit to emerge after a vulnerability is announced.
VentureBeat: How does AI support your defense capabilities?
Evans: Started integrating AI into your security playbook. By doing so, our security analysts now spend less time searching and hunting. AI is involved in Security Operations Center (SOC) products, conducting the initial triage analysis and saying, “Based on things we saw and things with my model, this is where I want to guide you.”
On the defense, we’re really beginning to see AI coming. Crowdstrike, Sentinel One, Microsoft Defender, and traditional enhanced detection and response (EDR) products used machine learning. However, AI enriches the ability to reach a high probability of identifying threats in EDR engines.
VentureBeat: What happens at night when it comes to AI and cybersecurity?
Evans: What worries me quite a bit is the deep fake. Read multiple stories about using deepfake to impersonate CEO and start wire transfers. They are of concern as they look very, very realistic.
But what is your biggest concern? The worst is one of those employees who retrieve customer data and put it in an AI engine that they don’t manage, and it’s useful for training your model.
VentureBeat: How did you explain this shadow AI risk to the board?
Evans: I remember when I had one of the first board meetings, they asked me. “Well, it’s an incredible productivity tool. But I don’t know how it can be used with employees. My biggest fear is when someone copy and paste customer data, or source code, which is our intellectual property.”
But I wasn’t just coming to the board with my concerns and issues. “Well, here’s my solution. I don’t want to stop people from being productive, but I want to protect it.” When I came to the board and explained how these enterprise browsers work, they said, “Okay, that makes a lot of sense, but can you really do it?”
VentureBeat: Learn about the island’s assessment and deployment process.
Evans: After the board meeting in October 2023, we began a rather long due diligence process. We’ve looked at some of the major vendors in the Enterprise Browser Space.
We will finally share why we finally went to the island. We needed to have control over the browsers people were using on their endpoints. It’s no use deploying an enterprise browser when someone is able to download and use Opera or “Frank’s Browser of the Month”;
Another reason we went on the island was really because of the speed of deployment. I remember calling the island sales representative and they said, “I believe we can roll this out to your company in a few weeks.” I said, “Oh, that’s BS.”
VentureBeat: But did they deliver?
Evans: They took it as a personal challenge! The island began rolling out with approximately 200 people in April 2024. First I went to the extended route. Expansion of the island of chrome and edges.
It wasn’t until July that the board asked, “What do you think?” And I said, “Why don’t I show you?” Murphy’s law demonstrations always fail, so I pulled up the screenshots. So I showed them a screenshot, “Here I’m in ChatGpt. I tried to paste something. I got the prompt: “Isle policy prevents you from doing this.”
They said, “Wow, this is great! But can people still use the tools to ask good questions?” I said, “Yeah, absolutely. They can’t put data into it.”
VentureBeat: Do you think the island will guarantee you and reduce the risk of ShadowAI?
Evans: It definitely helped us get the handle of Shadow AI. There is no 100% perfect security tool. With the island unfolding, we are definitely much less sleepy. If your employee is going to an AI instance that you don’t have a license, you can use it, but you can reasonably feel comfortable not being able to paste data or upload files.
It also helped me identify where the gaps were. Employees can find this really great AI widget, come to the security team and say, “Hey, look, check this out.” And we can go back to the product development team and understand how to enable this for our customers as well as our employees.
VentureBeat: How do you defend against deepfakes?
Evans: It’s difficult to wrap around your arms. There is a great security awareness program. Ask employees to use common sense. Do you really think CEO Sandeep Sahai will call you and ask him to buy an Apple gift card?
We set up a lot of checks and balances, like the two-person buddy check system. There is no technical solution for such things. It was human problems that had to implement human solutions.
VentureBeat: What advice would you give to other CISOs facing ShadowAI?
Evans: This is not just about blocks, it’s about enablements. It brings not only problems, but solutions. When I came to the board, I wasn’t just highlighting the risks. We proposed a solution that balances security and productivity.
Welcome to Shadow Ai Arms Race
Evans’ insights reveal how quickly Shadow AI has become an existential threat to all data-intensive businesses.
“We look at 50 new AI apps a day, and we already catalogue over 12,000 catalogs,” Itamar Golan, CEO of Propent Security, told VentureBeat, quantifying what the security team is calling, calling it the worst nightmare since ransomware.
Unauthorized use of AI and app onslaught have sparked fierce competition among security vendors. “Most traditional management tools don’t have comprehensive visibility into AI apps,” explains Vineet Arora, CTO at Winwire, to VentureBeat, identifying exactly why Shadow Ai thrives as the reason the Legacy Security architecture blinds it.
The Bender ecosystems crystallized into four different battlefields, each with weapons and weaknesses.
Enterprise Browser is the leading price. The most important of these is the island, which recently raised a $250 million funding round, a vote of trust from the investor community. While Island is betting on pre-cryptography visibility, Google Chrome Enterprise attacks Shadow AI differently, weaponizing market advantages and Google’s security stack. Chrome Enterprise Premium provides data loss prevention (DLP) controls that block data flows and other AI tools that block data flows, prevent cross-profile contamination, and perform real-time content scanning. The platform exposes Shadow AI usage patterns while blocking both accidental paste and intentional peeling. The strategic partnership with Zscaler and Cisco Secure Access amplifies Chrome’s reach and creates an ecosystem where zero trust principles are directly extended to AI interactions.
The SASE/SSE platform provides enterprise-scale defense. Netskope and Zscaler bring AI defense to shadows to scale through the Cloud-Native Security Access Service Edge (SASE) architecture. Both platforms process billions of transactions every day across global infrastructure, and Netskope specifically promotes its ability to monitor AI application usage across the enterprise. Their main limitation: When 73.8% of workplace ChatGPT usage occurs through personal accounts, SSL/TLS encryption forces the platform to prevent content inspections, forces it to rely on traffic patterns and metadata, leading to a gap in visibility where Shadow AI is not detected.
Traditional DLP vendors have a hard time adapting. Legacy vendors ForcePoint and Microsoft Purview have a strong legacy to trade when it comes to fighting Shadow AI. ForcePoint claims over 1,700 classifiers, and Purview leverages AI for triage tasks. But the problem is: Renovating the 20th century architecture for the threat of the 21st century. These platforms are great with compliance checkboxes and policy templates, but they can’t keep up with the faster pace of AI.
Like Daren Goeson, UEM’s Ivanti product management SVP told VentureBeat: Traditional DLP works at audit speed. Shadow AI moves at machine speed.
Special solutions fill important gaps. Innovation thrives in a niche that legacy vendors ignore. One example is Ivanti Neurons. This provides comprehensive device discovery through the UEM platform, exposing shadow AI hidden in endpoints that traditional tools have missed. Mike Riemer, Ivanti’s field Ciso, believes that “security experts will effectively utilize the capabilities of Gen AI to analyze the vast amount of data collected from a wide range of systems.” Dightal uses trans models to target developer teams, claiming twice the detection accuracy of API-based AI tools.
Comparison of Shadow AI Defense Solutions
Venture Beat Analysis
What is trying to move the market so quickly? A VentureBeat analysis shows that over 74,500 Shadow AI apps are actively deployed only to major consulting companies, increasing by 5% each month. By mid-2026, that number could reach 160,000. Each represents a potential data breaches, non-compliance, or a competitive intelligence leak.
Arora’s prescriptions are reduced through vendor hype. “Organisations need to define strategies with robust security, while ensuring employees can use AI technology effectively.
