LOLMIL: LIVING OFF THE LAND MODELS AND INFERENCE LIBRARIES

Summary from the TLDR InfoSec newsletter

PromptLock is a proof-of-concept malware developed by NYU researchers to operate autonomously using an LLM hosted by the researchers. The author of this post wanted to experiment with building a post-exploitation tool that leverages the LLM included in Microsoft's Copilot+ PCs to create malware that's not dependent on a Command and Control (C2) Server. By providing the malware with tools to identify, modify, and restart vulnerable services, they were able to exploit two services using the LLM in a test environment.