Lucene search
+L

1 matches found

Mend
Mend
•added 2026/05/16 1:49 a.m.•7 views

MAI-2024-0060

Large Language Models LLMs are susceptible to a sophisticated semantic mirror jailbreak attack. This attack utilizes a genetic algorithm to craft jailbreak prompts that closely resemble benign prompts in semantic terms, thereby circumventing defenses that rely on semantic similarity metrics. The...

6.3CVSS5.8AI score
SaveExploits0References1
Rows per page
Query Builder