Companies that use chatbots for customer service must always give people the option of speaking to a real employee, two Dutch regulators have said.

The Dutch privacy watchdog AP and competition authority ACM said organisations also need to make clear when customers are dealing with a chatbot, and ensure that the software does not provide evasive, misleading or incorrect information.

Current consumer law already requires companies to communicate with customers in a direct and effective way, but this is not always happening in practice, the agencies said on Thursday.

Rules for online platforms such as social media, marketplaces and other ‘intermediary services’ are clearer under the EU’s Digital Services Act, which requires a non-automated channel of communication.

The European Commission is preparing further changes to consumer law under the Digital Fairness Act to make these requirements explicit for other businesses as well.

From August 2, 2026, companies across the EU will also face new transparency requirements under the AI Act. These will require businesses to inform people when they are dealing with an automated system.

The AP and ACM said the Commission should go further by introducing rules to ensure AI chatbots are designed to be fair, recognisable and accessible, and that customers are not misled.

The regulators said complaints about chatbots are increasing rapidly. ACM research showed that the lack of human contact is one of the biggest frustrations for consumers, while chatbots often provide poor or even wrong answers. Many people also struggle to get through to a staff member if they want to.

The AP and ACM also warned of privacy and security risks. Chatbots are a form of generative AI trained on large amounts of data, which can include confidential information. This could make it easier for malicious actors to extract sensitive details, raising the risk of data breaches, they said.

Problems

In August, the ACM published research which showed two-thirds of Dutch consumers run into problems when using online platforms, ranging from blocked accounts to misleading algorithms.

Many users complained that their accounts or content were restricted without explanation, or that they could only reach automated helpdesks such as chatbots instead of a human contact point.