As new large language models, or LLMs, are rapidly developed and deployed, existing methods for evaluating their safety and discovering potential vulnerabilities quickly become outdated. To identify ...