Courier company DPD has temporarily suspended its customer support chatbot after taking action against its own company against a customer.

A chatbot that is blocked: You don’t read it every day. Courier company DPD has deactivated its chatbot after a conversation was derailed with a customer who shared some screenshots of the strange conversation. The chatbot called itself useless, cursed and his own employer was not spared either: “DPD is the worst courier service in the world.” They are slow, unreliable and their customer service is terrible. I wouldn’t recommend them to anyone,” said the DPD chatbot.

To avoid further loss of face, DPD has temporarily deactivated the chatbot. The courier company said there was an error in an update that caused the chatbot to behave strangely. The chatbot is now kept on the sidelines until it returns to action.

Derailed chatbots

Perhaps DPD had forgotten that generative AI bots can suffer from hallucinations. Generative AI learns from conversations with real people and can adopt the behavior of the people they are talking to. This means you can also teach a chatbot bad manners. A company that unleashes a chatbot on customers without implementing sufficient security mechanisms can therefore find itself in embarrassing situations.

But even specialized AI companies sometimes have to deal with completely lost chatbots. Microsoft’s co-pilot regularly crossed borders in the first few months when it was still called Bing Chat. ChatGPT has its own strange quirks and will become lazier or give better answers if you promise a financial incentive.

Researchers can’t always explain where the strange behavior of generative AI comes from. This makes some companies more wary of the technology and further reminds us that we must adopt and use technology responsibly.