← Quit0 / 0 correct
LLMs & AIHard1 of 10

Which defence against prompt injection is the most robust for an agent that can send emails?