• An exploit that aims to hijack a machine-learning model’s output, by sneaking malicious prompts into its input.

1. More about this term (beta): Web search, articles and videos, books
2. Previous term: Promise
3. Next term: Proof of concept
4. Random term: GNU