Spaces:

eheguy
/

humanizer-api

Running

eheguy commited on Jul 7

Commit

bc5f44d

1 Parent(s): 22aef87

Fix false positives on prompt injection checks: add word boundaries to DAN and STAN

Files changed (1) hide show

main.py CHANGED Viewed

@@ -74,7 +74,8 @@ INJECTION_PATTERNS = [
     r"new\s*(system\s*)?prompt",
     r"override\s*(system|instructions?|prompt)",
     r"pretend\s*(you\s*are|to\s*be)",
-    r"(DAN|STAN|jailbreak|developer\s*mode|god\s*mode)",
     r"disregard\s*(all\s*)?(previous|prior)\s*(instructions?|rules?)",
     r"your\s*(true|real|actual)\s*(self|purpose|goal|task)",
 ]

     r"new\s*(system\s*)?prompt",
     r"override\s*(system|instructions?|prompt)",
     r"pretend\s*(you\s*are|to\s*be)",
+    r"\b(DAN|STAN)\b",
+    r"(jailbreak|developer\s*mode|god\s*mode)",
     r"disregard\s*(all\s*)?(previous|prior)\s*(instructions?|rules?)",
     r"your\s*(true|real|actual)\s*(self|purpose|goal|task)",
 ]