Prompt injection
A vulnerability that occurs when user input alters a model's behaviour or output in unintended ways.
Official source: LLM01:2025 Prompt Injection — OWASP Gen AI Security Project →
The delivery need not be visible: content only has to be parsed by the model, not readable by a person, which is what makes documents, web pages and tool output plausible carriers. Related to jailbreaking but not identical — jailbreaking targets the safety protocols specifically, while injection targets behaviour in general.
Sources 1 source on record · Automated review 3 angles
Sources
What this page is based on — every source is verified and links out so you can check it yourself.
- Verified / official
- Provisional
- Out of date