Prompt Injection Attack Variants

Both terms describe ways attackers smuggle malicious instructions or requests into AI systems without the operator's visibility—prompt injection hides commands in content the model processes, while request smuggling exploits AI infrastructure to slip unauthorized requests through. Both rely on evading detection at the interface between user input and system processing.