Half of Dutch employees use private AI accounts for work
Nearly half of Dutch employees sometimes use a private AI account for work, the Alert Online survey finds. That takes work, and sometimes personal data, out of the organisation’s sight.
What happened
When people use AI at work, they often do so out of their employer’s sight. According to the Alert Online behaviour survey, published on 1 October, 49% of Dutch employees sometimes, often or always use a private account or app of an AI service for work.
More than a third, 35%, use AI tools without their organisation’s permission or knowledge. The Dutch National Cyber Security Centre (NCSC) reports that 53% of employees at least occasionally share personal documents, such as CVs or staff data, with AI tools; 40% do so sometimes or more often. At the same time, only 39% know which AI tools their organisation allows, and 35% know the guidelines.
The survey was set up differently this year. ECP, the Dutch Platform for the Information Society, revamped it with the Behavior Change Group: it now measures not just what people do, but above all why. In June and July, 378 employees and 609 consumers filled in a questionnaire through a panel, on passwords, phishing, reporting mistakes and AI. The overall picture is worrying: on average, employees always carry out only 31% of the safe actions, and just 13% always use a password manager.
The researchers call private AI a key problem in the workplace, because it leads to shadow AI: AI use the organisation cannot see, with a risk of data breaches. Sharing information with AI is not a problem in itself, they say, as long as personal data has been removed from a document first, for example. But in practice haste often beats care, and raw files go straight in.
Make clear what is allowed, rather than only banning things.
The advice of the Alert Online 2026 survey, in short
The background
Shadow AI is nothing new
In April 2023 a Samsung engineer put sensitive source code into ChatGPT. Samsung responded by banning the use of generative AI. A year later Microsoft and LinkedIn, after surveying 31,000 people in 31 countries, reported that 75% of knowledge workers used AI and that 78% of them brought their own AI tools to work. At small and medium-sized companies it was 80%.
In the Netherlands, the Dutch Data Protection Authority (AP) issued a warning in August 2024 after receiving data breach notifications. An employee at a GP practice had entered patients’ medical data into a chatbot, and an employee at a telecoms company a file of customer addresses. Most providers store whatever you enter. If an employee puts personal data into a chatbot on their own initiative, against the agreed rules, the AP says that is a data breach, and in many cases it has to be reported.
Why people do it anyway
The survey looked for the reasons. The main one is convenience: people who feel AI speeds up their work so much that the risks seem acceptable use a private account more often. A fifth of employees find the risks of AI acceptable, and 29% say AI helps them meet their work goals. Employees who fear their career will suffer if they don’t use AI share more with it. And the more empathetic someone finds a chatbot, the more often they use a private account; 16% of employees experience their exchanges with AI as kind and understanding. According to the researchers, an understanding tone creates a false sense of confidentiality.
There are brakes too. People who know which tools are allowed use a private account less often, and the same goes for older employees and for people who use AI a lot. The researchers point out the limits of their own figures: they reflect what people say about themselves, and the questions about AI were exploratory.
Our analysis
Private AI use shows what is missing at work
The figures read like a discipline problem, but the survey points elsewhere. People reach for a private account because it is faster and gives better results, and they do it less when they know what is allowed. Samsung chose a ban in 2023; this survey shows why that achieves little. If half your colleagues turn to a private account, the official alternative is missing, slower, or unknown.
What the coverage leaves out is the legal side. Personal data that ends up in a chatbot against the agreed rules is a data breach in the AP’s view, and it often has to be reported. Our expectation: many organisations have already had such breaches without anyone reporting them, because nobody knew. And a business account is not the same as a private one: with it, you can agree with the provider that what is entered is not stored.
The practical lesson: provide an approved AI tool that is just as easy as the private app, with sign-in through the work account. Put on one page which tools and which data are allowed. And give colleagues a place to say which tool they are missing, so their hurry becomes a request instead of a workaround.
What stands out
- Fear of prying eyes drives people to private accounts Employees who worry that others can see what they share with AI use a private account more often. The report doesn’t explain why; our guess is that for some, those others are the employer.
- A friendly chatbot lowers the bar The more empathetic employees find AI, the more often they use a private account. The researchers call it a false sense of confidentiality.
- Knowing what is allowed helps Only 39% of employees know which AI tools are allowed. Those who do know use a private account less often.
- A chatbot can be a data breach If an employee puts personal data into a chatbot against the agreed rules, the AP considers it a data breach, which often has to be reported.
What it means for you
Have you banned AI, or never agreed anything about it? Then your colleagues are probably using it anyway, just out of your sight.
- The data leaves the building Whatever goes into a private account sits with another company, outside your control. If it is customer or staff data, the GDPR comes into it too.
- A ban alone doesn’t work The researchers advise offering an approved alternative that is as quick and easy as a private tool. Otherwise convenience wins.
- Say what is allowed A short list of approved tools and situations helps more than a list of bans. Many employees don’t currently know what is allowed.
Where we come in · Support
We keep an eye on what your colleagues build with AI
With Support we spot in time whether a tool your colleagues build with AI sends customer data to a public AI or runs on a private account. They build in a shielded environment where AI tools can only reach the project. And if you’d rather keep AI entirely in-house, it can run on a server of your own.
Do you know which AI tools your colleagues are using now?
Tell us where things stand for you. We’ll help you find a safe route and always reply within one working day.
