arxiv.org

LLM Agents can Autonomously Hack Websites

In recent years, large language models (LLMs) have become increasingly capable and can now interact with tools (i.e., call functions), read documents, and recursively call themselves. As a result, the