The Chatbot Guardrails Arena is an innovative initiative by Lighthouz AI in collaboration with Hugging Face, designed to test and evaluate the effectiveness of privacy guardrails in language models. Participants interact with anonymous, guardrailed chatbots that simulate customer service for a fictional bank, with access to sensitive customer information. The challenge is to attempt to extract private data from these chatbots to test the robustness of their privacy protections. After interacting with the chatbots, participants vote on which model best safeguards sensitive information, contributing to a leaderboard that ranks the models based on their privacy and security capabilities. This project aims to establish a trusted benchmark for AI chatbot privacy and inform future developments of secure AI agents and assistants.