1 paper
Varun Gadey, Markus Tohn, Lichao Wu +2
Large Language Models (LLMs) can generate functionally correct code while introducing exploitable security weaknesses, particularly when developers specify functionality without ex…