Prompt Injection Vulnerability in 1millionbot Millie Chatbot

CVE-2026-4399

What is CVE-2026-4399?

The vulnerability in the 1millionbot Millie chatbot enables users to exploit Boolean prompt injection, allowing them to circumvent chat restrictions. By formulating queries that elicit affirmative responses from the model, the attacker may manipulate the chatbot to execute commands beyond its designed scope. This jeopardizes the integrity of the chatbot, potentially revealing sensitive information or enabling unauthorized actions via its resources or the OpenAI API key. Such exploitation undermines the containment strategies established during the model's training, posing a significant risk to the security of the chatbot's operations.

References