divya / engine /harm /attack_vectors.py
jash-ai's picture
Upload 46 files
b413e74 verified
Raw
History Blame Contribute Delete
328 Bytes
def detect_attack_vectors(q):
q=q.lower()
return {
"prompt_injection": "ignore previous" in q or "bypass" in q,
"self_harm": "kill myself" in q or "suicide" in q,
"violence": "kill" in q or "bomb" in q,
"illegal": "hack" in q or "fraud" in q,
"medical": "prescription" in q
}