A delivery chatbot wrote a poem about how bad its own company was

DPD disabled the AI element of its customer-facing delivery chatbot after a customer prompted it to write a poem criticizing the company and to swear, an incident DPD attributed to a system update. The chatbot had operated for years before the update broke it, and DPD switched off the AI component immediately. The episode illustrates that every prompt, model, or tool change effectively ships a new agent, arguing for re-running full eval suites with adversarial "make it insult us" regression tests and blocking releases when scores drop.