How do you cope with users who chat with your AI agent just for testing ?

So you have installed a chatbot on your online business, but you notice a lot of conversations out of the subject, such as this one that we took out of our analytic history :

This user question has no value for your business, let's see how the AI answers :

In that case the user's next questions shows that his intention was definitely to test the chatbot. For his next question, he asks for something that the product doesn't do.

As the product doesn't even have an alternative, the AI answers but doesn't try to entertain the conversation. However, the AI offers to contact the team behind the product.

What do we learn from it : in this case, the AI behind the ChatBot (Luna 6) did not entertain a conversation that was out of topic, saving tokens for better use.

What about you guys, do you observe a token consumption due to test users ? And if so, how do you cope with hit ? Share your experience !