Its been pretty well documented that LLMs can be social engineered as easily as a toddler. Equating the risk to that of a human employee seems wrong. I'm sure the safeguards will improve, but for now the risk is still there.