Skip to content

Glossary: Modern Vulnerabilities, Malware and Physical Attacks

Prompt injection

An attack on a large language model in which text written by an attacker changes what the model does, for example by overriding a chatbot's rules. It is the top weakness in the OWASP Top 10 for LLM Applications, and it is one reason an AI chatbot adds to the attack surface.

Where the book teaches it

Related terms

Large language model (LLM)OWASP Top 10 for LLM Applications