AI expands our capabilities, but also our mistakes

At our Cybersecurity Day, we asked a simple question: what business data should never end up in an AI tool? The answer seems obvious. Yet in practice, things regularly go wrong – and with increasing frequency. According to the Verizon Data Breach Investigations Report 2026 Today, 45% of staff regularly use AI on their work devices – three times as many as a year ago. Two in three do so using a personal account, outside the remit of IT. Well-known examples also highlight the risk: Samsung banned ChatGPT after staff had shared sensitive source code. At Eindhoven City Council, a large amount of personal data ended up in public AI tools. Usually, there is no malicious intent behind this. An employee simply wants to quickly convert an Excel file, have a text rewritten or merge a PDF. But by using a free tool and a personal account, that data is effectively handed over to third parties.

Three questions to ask yourself

1. What business data should never end up in AI? And do you know which AI tools your staff are using today?
2. Do you keep track of how and where your data moves? Can your current toolset actually monitor that?
3. Do you still grant full network access via VPN? And how does that affect the user experience?

Are you unsure about any of these questions? If so, you’re certainly not alone.

Users, data and apps are everywhere

In the past, everything was protected by a firewall at head office. Today, staff work from home, whilst travelling and at other branches, using dozens of SaaS apps, social media and AI tools. To secure this, more and more standalone solutions were introduced: SD-WAN, ZTNA, CASB, web filtering, DLP. Often from different suppliers, leading to overlap, conflicts and high licence costs. And one question remains: who has the full picture?

A personal bodyguard for every user

Imagine bringing all those tools together in a single personal guardian that accompanies every user, wherever they work. A guardian that:

• knows and applies all the rules;
• follows each step;
• operates quickly and close to the user;
• adjusts rather than blocking everything;
• protects both users and data;
• is always available, up to date and active.

That’s what Managed AI CloudGuard does. The platform analyses prompts and attachments before they even reach an AI tool. If an employee attempts to paste sensitive data into a public AI tool, they will, for example, receive a warning, be required to justify their decision, or be redirected to an approved tool. This is how you make AI as safe as possible, rather than banning it.

And it goes beyond AI alone. CloudGuard can distinguish between the business version of OneDrive and a personal account, ensuring that work data isn’t transferred to a private account without being noticed. It also monitors the device itself: sensitive data transferred to a USB stick or printer is equally subject to your rules.

“The explosive growth in AI use is causing concern amongst business leaders: ‘That must certainly not be allowed in ChatGPT!’ But the reality is that employees have been leaking sensitive data for years via thousands of free or unknown cloud applications and shadow IT.”

Roel Van Looy, Technical Director at VanRoey

How does it work?

Managed AI CloudGuard is built on Netskope, a Leader in the Gartner Magic Quadrant for both SASE and SSE. Everything happens in the cloud:

• One console and one set of rules: simpler and cheaper than a setup involving different suppliers.
• A light-coloured police officer on your staff’s devices, easily rolled out via Microsoft Intune.
• More than 120 data centres worldwide, with direct links to the major cloud platforms. That way, everything runs smoothly.
• Recognises more than 86,000 SaaS apps, including AI tools.
• Local processing in European data centres, in accordance with the GDPR and NIS2.
• A modern alternative to your VPN: secure access to internal applications, without full network access.

Curious to find out how that works from a technical point of view? On our page about Managed AI CloudGuard You can read all about how it works, the role of Netskope and how it fits within the GDPR and NIS2.

Why managed?

A powerful platform is one thing. Configuring and maintaining it properly is quite another. Rules that are too strict lead to frustration and even downtime. Rules that are too lax result in data leaks. That is why our certified cybersecurity specialists help to manage the platform:

• advice and a phased roll-out;
• a tried-and-tested basic configuration;
• quarterly reviews with feedback;
• no configuration drift;
• maintenance of the components required for secure access to your internal applications.

You can optionally link CloudGuard to our SOC. We then monitor the logs and alerts from your entire network security infrastructure, including SaaS and AI usage. As a Netskope Select Tier Service Delivery Partner, we have the necessary expertise and experience to do this.

Find out for yourself with a free PoC

Curious to find out which AI and cloud apps are already being used in your organisation, and where the risks lie? With our free AI Security PoC, we’ll map this out over the course of a week. It’s not a major project, but it will give you a clear picture to work from. We won’t block anything during the PoC. Request your free PoC or leave your details below to arrange a no-obligation consultation.