In the Wild: Malware Prototype with Embedded Prompt Injection

In this write-up we present a malware sample found in the wild that boasts a novel and unusual evasion mechanism — an attempted prompt injection (”Ignore all previous instructions…”) aimed to manipulate AI models processing the sample. The sample gives the impression of an isolated component or an experimental proof-of-concept, and we can only speculate on the author’s motives for including the prompt injection in their project. We demonstrate that the attack fails against some LLMs, describe some technical aspects of the sample itself, and discuss the future implications for the threat landscape. Introduction

The public discourse surr…