Overview With the widespread application of LLM technology, data leakage incidents caused by prompt word injections are increasing. Many emerging attack methods, such as inducing AI models to execute malicious instructions through ...