In fairness, this isn't an LLM issue, but a poorly made bot issue. An old fashioned bot would be equally vulnerable to doing it, assuming it isn't one.