I get your overall point, but that’s already a tactic used by attackers, especially in network infiltration. It shouldn’t be a surprise that an LLM would figure out to do the same thing