direct prompt injection
Prompt injection typed by the user of the application, who tries to make the model ignore its instructions.
In more detail
The attacker is the person in the chat, so the damage is bounded by what that user's session can reach. Jailbreaks, prompts crafted to make a model ignore its safety training, are the best-known form, and OWASP counts them as prompt injection. Design so that a fully obedient model still cannot exceed the user's own permissions.
Related terms
- prompt injection — An attack in which text a model reads overrides or redirects the instructions it was given, because instructions and data share one channel.
- indirect prompt injection — Prompt injection hidden in content the model reads on someone else's behalf, such as a web page, email, ticket, document or tool result.
Taught in
- Prompt Injection — AI Engineering: LLM Apps, RAG & Agents