prompt injection
Prompt injection is an attack that places malicious instructions in an AI model's input to redirect its behavior.
Prompt injection is an attack that embeds instructions in material processed by an AI model to make it depart from its intended task. Unlike an ordinary prompt that asks the model to perform a task, it aims to make the model obey instructions in content it should not trust.
When the instructions are hidden in material the model retrieves or reads, such as a webpage, document or search result, the method is called indirect prompt injection.
This entry is based on AIPOST articles and widely known facts. If something is wrong, please send us a correction request.
Articles covering this entry
These monitors look for prompt injection, an attack that hides instructions in outside content to hijack an AI, stop high-risk autonomous actions and step in be…
Developers often ask how to protect agents from leaked credentials and prompt injection, an attack that hides instructions in content the model reads.
It filters prompt injection, where an attacker slips in instructions to hijack the model, along with jailbreak attempts and leaks of sensitive data.
Since the real key never exists inside the container, an agent that has been fooled by prompt injection still has nothing valuable to leak.
Claude Opus 5.5 has improved resistance to direct prompt injection, but a clear boundary remains useful.
Indirect prompt injection places instructions in material an AI system reads as data, rather than in a direct request from its user.
Those checks included prompt injection, where instructions embedded in material a model processes try to redirect its behavior, and deliberately impossible task…