Collected research
Exploring Prompt Injection Attacks
Explores prompt injection against model-backed applications through imperative instructions and attacker-supplied examples. Contrasting failed and successful payload variants, delimiters and retry behavior shows how input presented as data can redirect a model’s task or expose instructions from its original prompt.
Record
- Researcher
- Jose Selvi
- Published by
- NCC Group
- Topic
- Injection
In the archive
Related sources
- Prompt injection attacks against GPT-3
- Declassifying the May 2022 prompt-injection disclosure
- Riley Goodside original prompt-injection demonstration
- I don’t know how to solve prompt injection
- Melon Goose prompt-extraction payload
- Simon Willison prompt-extraction demonstration
- Yohei Nakajima injection-test proposal
Tags
This page is the archive's own catalogue record. The research is the work of Jose Selvi, first published at the original source. Preserved copies are kept so the citation survives its host; this one was last captured on .