Prompt Injection
Prompt injection is an attack that hides malicious instructions in text an AI model reads, causing it to ignore its intended instructions or take unwanted actions.
What is Prompt Injection?
Injection can be direct, from the user, or indirect, hidden in web pages, emails or documents processed by an AI agent. It is the top risk on the OWASP list for LLM applications and is hard to fully prevent.
How it works
Defences include separating trusted and untrusted content, limiting tool permissions and requiring confirmation for sensitive actions.
← All Artificial Intelligence & Superintelligence definitions · ← Back to the A–Z definitions