TABLE OF CONTENTS (contd)

Explore the top 10 security vulnerabilities affecting Large Language Models. Click on any topic to learn more or use the navigation below to read from start to finish.

06

Excessive Agency

When AI systems gain too much freedom to act autonomously

07

System Prompt Leakage

When hidden system instructions are revealed to attackers

08

Vector and Embedding Weaknesses

Vulnerabilities in how AI systems store and retrieve information

09

Misinformation

When AI confidently presents false information as truth

10

Unbounded Consumption

When AI systems use excessive resources without proper limits