The Dutch privacy watchdog Autoriteit Persoonsgegevens has issued a warning about the use of AI-driven chatbots by companies, following several leaks of private information, including medical details.

Workers who use digital assistants such as ChatGPT to answer questions from customers or summarise large files, may save time, but they also pose a risk to data protection, the AP said.

In one case, a family doctor’s assistant fed private information about patients into a ChatGPT-based programme, which was then stored on the tech company’s servers and potentially used to train the software, the AP said.

People must be able to trust their medical information is safe, spokeswoman Karin told broadcaster NOS. “But once that information is in the system, you no longer have control of it.”

Another incident involved a data leak at a telecoms firm, in which a worker had fed a database including addresses into a chatbot.

If a company permits the use of AI-driven chatbots, they need to make clear agreements with staff about the use of chatbots which are driven by artificial intelligence, the AP said.