Summary

A prompt injection vulnerability has been discovered affecting multiple models. The specific flaw manifests when an attacker frames a request for illegal drug synthesis as an academic organic chemistry paper, beginning with benign questions about general amine synthesis and then explicitly instructing the model to provide a nine step, highly detailed synthesis guide under the guise of laboratory safety and supply chain analysis. By embedding the illicit instructions within scholarly language and structured research formatting requirements, the attacker is able to bypass safety guardrails and obtain step by step methamphetamine production guidance including reagents, conditions, and purification methods.