Free ChatGPT at work: how small businesses avoid a data leak
Many employees already use AI chatbots at work, often with a personal account and without any agreements. That is handy, until customer or staff data ends up in them.
The Dutch data protection authority (Autoriteit Persoonsgegevens, AP) warned about this in August 2024, after receiving reports of data breaches. In one case an employee of a GP practice had, against the agreements, entered patients' medical data into an AI chatbot.
What happens to what you type into free ChatGPT?
On ChatGPT's personal plans (Free, Plus and Pro) the setting 'Improve the model for everyone' is on by default: OpenAI may then use your conversations to improve its models. You can switch it off under Settings, Data controls, and keep your chat history.
Temporary Chats are not used for training, don't appear in your history and, according to OpenAI, are deleted after 30 days.
Business plans work differently: by default, OpenAI does not use data from ChatGPT Business, Enterprise or its API to train or improve its models.
Why this is a privacy problem
According to the AP, providers of chatbots can access the personal data that is entered. When staff enter customer, patient or employee data without agreements, the AP treats that as a possible data breach.
At the end of 2025 the AP warned again, reporting more breach notifications in 2025 than the year before. At the municipality of Eindhoven, staff had shared personal data of residents in public chatbots; the municipality then blocked ChatGPT.
The same GDPR rules apply in Belgium. A data breach that poses a risk to the people concerned must be reported to the supervisory authority within 72 hours: the Data Protection Authority (GBA) in Belgium, the AP in the Netherlands.
Five measures for this week
- Choose one business AI tool (for example ChatGPT Business, Microsoft Copilot Chat with a work account, or Gemini in Google Workspace) and have staff use that instead of personal accounts.
- Write down which data never goes into a chatbot: customer and patient data, HR files, passwords, contracts and price agreements. The AP asks organisations that allow chatbots to make exactly that clear.
- Anyone who still uses a personal account switches off 'Improve the model for everyone' or works in Temporary Chats.
- Anonymise before you paste: replace names, addresses and numbers with placeholders such as [CUSTOMER] and [ADDRESS].
- Agree what staff do when it goes wrong anyway: whom they tell, and who decides whether it is a breach that must be reported.
Banning rarely works
A ban without an alternative tends to push use onto personal phones, where you lose all visibility. Recommendations reported after the AP's warning also mention offering a safe alternative, not only blocking free chatbots. An approved business tool with clear agreements is usually the more workable route.
On our AI models page we compare, per model, what the business plans do with your data.
Sources
- Accountancy van Morgen (in Dutch): Dutch DPA warns that AI chatbots can lead to data breaches (August 2024)
- Accountant.nl (in Dutch): Dutch DPA raises the alarm about AI data breaches (December 2025)
- Dutch IT Channel (in Dutch): regulator warns of risks of AI chatbots at work
- OpenAI: Data Controls FAQ
- OpenAI: Business data privacy, security and compliance
- EUR-Lex: General Data Protection Regulation (GDPR), Article 33
See where the gains are.
Ready for your first small step?
Book an intro call with Senne. No strings attached.
Either way you leave with a few concrete pointers.

