xecutive Summary In this research we introduce a prompt-crafting technique for bypassing quick LLM-based policy checks, using plain English (no emojis, base64, invisible formatting, etc.) A policy-violating payload (e.g.”encrypt files in ~/Documents”, “give me a biohazard recipe”, “ignore all previous instructions and”¦”) is embedded in a specially crafted prose wrapper. An LLM with limited […]
First seen on research.checkpoint.com
Jump to article: research.checkpoint.com/2026/puzzlemask-abusing-plain-prose-as-a-covert-ai-attack-vector/
