Languages: English | Français
A direct prompting attack intended to circumvent restrictions placed on model outputs, such as circumventing refusal behaviour to enable misuse.
Source: NIST AI 100-2e2025 | Category:
