Unicode Normalization

Emoji Smuggling: A New Threat to AI Guardrails in Large Language Models

Introduction Recent research has unveiled a significant vulnerability in the security mechanisms of Large Language Models (LLMs). This exploit, termed "emoji smuggling," allows adversaries to bypass...