Prompt Injection is an attack in which malicious or unintended instructions attempt to manipulate a language model into ignoring its intended behavior or revealing restricted information.
Voice AI developers implement defenses against Prompt Injection to protect AI agents, enterprise knowledge bases, tool access, and sensitive customer data from malicious prompts or indirect instructions.