It’s pretty clear to me this is an intentional choice made by the openai team to collect a corpus of common methods to try and jailbreak an conversational AI so that future models can learn to avoid these methods in the future.
Never ascribe to malice what can be ascribed to stupidity works for genius as well.