Great article, but doesn't talk about the potentially _most_ dangerous form of mistakes: an adversarial LLM trying to inject vulnerabilities. I expect this to b

Great article, but doesn't talk about the potentially _most_ dangerous form of mistakes: an adversarial LLM trying to inject vulnerabilities. I expect this to become a vector soon as people figure out ways to accomplish this