Yet again, simply asking an LLM to be naughty in the right way causes it to be naughty, and yet we still trust them with our code and data
As opposed to humans, who are immune to social engineering.
As opposed to humans, who are immune to social engineering.