Indirect prompt injection lets attackers bypass LLM supervisor agents by hiding malicious instructions in profile fields and contextual data. Learn how this attack works and how to defend against it.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results