What is Prompt Injection?
Prompt injection is an attack where malicious instructions hidden in content (a web page, document or email) trick an AI system into ignoring its rules or leaking data. It is the top security risk for LLM-powered applications.
Because LLMs follow instructions in the text they read, attacker-controlled content can hijack an agent — exfiltrating data or triggering unintended actions. Defences include treating retrieved content as untrusted data, least-privilege tool access, output validation and human confirmation for sensitive actions.