Morning Edition · Friday, July 31, 2026Published at 2:00 AM EDT · New York

New Jailbreak Method Uses Dual-Layer Encoding to Slip Past LLM Moderation

RoguePrompt hides instructions in a self-reconstructing encoding that safety filters do not parse, then has the model decode and execute them.

Researchers describe RoguePrompt, a method that circumvents large-language-model (LLM) moderation by wrapping malicious instructions in a dual-layer encoding the model reconstructs and executes at inference time. The attack targets the gap…