| 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 | |
| } | |