A method for preventing a malfunction of a filling device when a container is filled by a user with a beverage. The method includes providing a filling device comprising a control unit which controls the filling device, a camera which takes an image of a container currently being used with the filling device and which outputs the image to a classifier which then uses a trained learning algorithm to analyze the output of the image of the container provided by the camera. The trained learning algorithm of the classifier analyzes which container is currently being used based on characteristics of the container so as to classify the container into a predefined class. The predefined class is then employed by the control unit to prevent the malfunction of the filling device.