Consider the following interaction. Simple educational prompts now seem to trigger the ethical guardrails.
I also tried with different and clearer variations of the above prompt in fresh chat contexts (in case of an overlooked ambiguity in the original prompt). Results were the same.